feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
"""FastAPI router for the dashboard backend."""
|
|
|
|
|
|
|
|
|
|
from __future__ import annotations
|
|
|
|
|
|
|
|
|
|
import hmac
|
|
|
|
|
import logging
|
|
|
|
|
import os
|
|
|
|
|
from typing import Any
|
|
|
|
|
|
|
|
|
|
import httpx
|
|
|
|
|
from fastapi import APIRouter, Depends, HTTPException, Request
|
|
|
|
|
from fastapi.responses import RedirectResponse, Response
|
feat: restructure Open SWE Review tab + wire create_prs (#1319)
* feat(dashboard): restructure Open SWE Review tab + wire create_prs
Restructures the dashboard around two related changes the reviewer settings
have been asking for:
- Wire profile.create_prs. Defaults to true (opt-out); when off the system
prompt gets a `Pull Request Policy Override` section telling the agent
to push the branch and notify with the branch URL instead of opening a
PR. Removes the noop Slack Notifications / Allow Artifacts / First Name
/ Last Name controls and their schema fields.
- Repositories opt-in for Open SWE Review. New per-team enabled list
stored in the LangGraph Store (`["enabled_review_repos"]`). Every
reviewer webhook chokepoint now goes through `_is_repo_enabled_for_review`
which AND-combines the existing env allowlist with the dashboard list.
Default is empty (opt-in) — admins enable repos per-installation from
the new Repositories page nested under Open SWE Review.
- Open SWE Review tab now mirrors the Cursor "rules" pattern: main page
shows installation rows + a Rules entry; both drill into nested pages
(/review/repositories/$owner and /review/styles) with a back link.
- Adds the new logo/favicon assets shipped from sidebar + html head.
Tests pass with a new autouse fixture (`tests/conftest.py`) that defaults
`is_review_repo_enabled` to True for existing allowlist tests.
* fix(dashboard): make main content scroll independently of the sidebar
Outer flex container was min-h-svh, so it grew with main's content and the
whole page scrolled — sidebar moved with it. Pin to h-svh + overflow-hidden
so the sidebar stays put and only <main> scrolls.
* fix(dashboard): make disabled repo toggles obviously disabled
Switch's disabled state used opacity-50 against a muted background, so
the not-admin state looked nearly identical to the off state. Bump to
opacity-40 + grayscale, and wrap each repo toggle in a span carrying a
native hover tooltip explaining why it's disabled.
* fix(switch): handle base-ui's data-disabled state
base-ui's Switch.Root sets data-disabled (not the HTML disabled attribute)
when disabled, so Tailwind's disabled: variant never matches and the
button keeps its cursor-pointer + clickable look. Mirror the styling
under the data-[disabled] variant and add pointer-events-none so the
disabled state is both visible and actually unclickable.
* feat(dashboard): paginate per-installation repository list
20 repos per page with Prev / page X of Y / Next controls at the bottom.
Pager only renders when there are more than 20 repos. Page resets to 0
when navigating between installations.
* feat(dashboard): global default model selectors for Agent + Reviewer
Adds team-wide default model + reasoning effort for both agents in the
Admin tab so operators can switch models without redeploying.
Resolution chain:
Agent: hardcoded -> LLM_MODEL_ID env -> team default -> user profile
Reviewer: hardcoded -> LLM_MODEL_ID env -> team default -> per-call configurable
Team defaults live in team_settings and are validated against the
SUPPORTED_MODELS allowlist + the model's supported reasoning efforts.
'Inherit from env' clears the override and falls back to LLM_MODEL_ID.
* refactor(models): drop LLM_MODEL_ID env in favour of the team default
The team default is now the single source of truth for the runtime model
choice; per-user (agent) and per-call configurable (reviewer) selections
still win on top. When no admin has touched the team default, it surfaces
the hardcoded fallback (DEFAULT_MODEL_ID + its default effort), so the
admin UI's dropdown is always pre-populated with a sensible value.
The Admin UI loses the 'Inherit from env' option since there is no longer
an env layer to inherit from.
* chore(models): set hardcoded fallback to gpt-5.5 medium
Decouple the team-default boot value (gpt-5.5 / medium) from each model's
ProfileForm-suggested default_effort so we can change one without nudging
the other. The Opus xhigh default for new user profiles is unchanged.
* feat(dashboard): trigger-mode copy, Coming Soon badges, logout in My Settings
- Rename trigger mode 'ready_for_review' -> 'once_per_pr' with new
description copy that matches the screenshot. Legacy stored values
fall back to 'every_push' on read so the UI never shows an unknown
selection.
- Add a 'Coming soon' badge + greyed-out + disabled state on the
controls that don't have runtime consumers yet: Trigger Mode,
Autofix Mode, Autofix Severity Threshold, and Automatically fix CI
failures. SettingsRow grew a comingSoon prop to keep this consistent.
- My Settings drops the noop PR Preferences section and adds a Sign
Out button. preferred_pr_destination is removed from the profile
schema; old records get the field popped on next write.
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
2026-05-21 09:17:07 -07:00
|
|
|
from pydantic import BaseModel
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
|
|
|
|
|
from .admin import is_admin
|
feat: restructure Open SWE Review tab + wire create_prs (#1319)
* feat(dashboard): restructure Open SWE Review tab + wire create_prs
Restructures the dashboard around two related changes the reviewer settings
have been asking for:
- Wire profile.create_prs. Defaults to true (opt-out); when off the system
prompt gets a `Pull Request Policy Override` section telling the agent
to push the branch and notify with the branch URL instead of opening a
PR. Removes the noop Slack Notifications / Allow Artifacts / First Name
/ Last Name controls and their schema fields.
- Repositories opt-in for Open SWE Review. New per-team enabled list
stored in the LangGraph Store (`["enabled_review_repos"]`). Every
reviewer webhook chokepoint now goes through `_is_repo_enabled_for_review`
which AND-combines the existing env allowlist with the dashboard list.
Default is empty (opt-in) — admins enable repos per-installation from
the new Repositories page nested under Open SWE Review.
- Open SWE Review tab now mirrors the Cursor "rules" pattern: main page
shows installation rows + a Rules entry; both drill into nested pages
(/review/repositories/$owner and /review/styles) with a back link.
- Adds the new logo/favicon assets shipped from sidebar + html head.
Tests pass with a new autouse fixture (`tests/conftest.py`) that defaults
`is_review_repo_enabled` to True for existing allowlist tests.
* fix(dashboard): make main content scroll independently of the sidebar
Outer flex container was min-h-svh, so it grew with main's content and the
whole page scrolled — sidebar moved with it. Pin to h-svh + overflow-hidden
so the sidebar stays put and only <main> scrolls.
* fix(dashboard): make disabled repo toggles obviously disabled
Switch's disabled state used opacity-50 against a muted background, so
the not-admin state looked nearly identical to the off state. Bump to
opacity-40 + grayscale, and wrap each repo toggle in a span carrying a
native hover tooltip explaining why it's disabled.
* fix(switch): handle base-ui's data-disabled state
base-ui's Switch.Root sets data-disabled (not the HTML disabled attribute)
when disabled, so Tailwind's disabled: variant never matches and the
button keeps its cursor-pointer + clickable look. Mirror the styling
under the data-[disabled] variant and add pointer-events-none so the
disabled state is both visible and actually unclickable.
* feat(dashboard): paginate per-installation repository list
20 repos per page with Prev / page X of Y / Next controls at the bottom.
Pager only renders when there are more than 20 repos. Page resets to 0
when navigating between installations.
* feat(dashboard): global default model selectors for Agent + Reviewer
Adds team-wide default model + reasoning effort for both agents in the
Admin tab so operators can switch models without redeploying.
Resolution chain:
Agent: hardcoded -> LLM_MODEL_ID env -> team default -> user profile
Reviewer: hardcoded -> LLM_MODEL_ID env -> team default -> per-call configurable
Team defaults live in team_settings and are validated against the
SUPPORTED_MODELS allowlist + the model's supported reasoning efforts.
'Inherit from env' clears the override and falls back to LLM_MODEL_ID.
* refactor(models): drop LLM_MODEL_ID env in favour of the team default
The team default is now the single source of truth for the runtime model
choice; per-user (agent) and per-call configurable (reviewer) selections
still win on top. When no admin has touched the team default, it surfaces
the hardcoded fallback (DEFAULT_MODEL_ID + its default effort), so the
admin UI's dropdown is always pre-populated with a sensible value.
The Admin UI loses the 'Inherit from env' option since there is no longer
an env layer to inherit from.
* chore(models): set hardcoded fallback to gpt-5.5 medium
Decouple the team-default boot value (gpt-5.5 / medium) from each model's
ProfileForm-suggested default_effort so we can change one without nudging
the other. The Opus xhigh default for new user profiles is unchanged.
* feat(dashboard): trigger-mode copy, Coming Soon badges, logout in My Settings
- Rename trigger mode 'ready_for_review' -> 'once_per_pr' with new
description copy that matches the screenshot. Legacy stored values
fall back to 'every_push' on read so the UI never shows an unknown
selection.
- Add a 'Coming soon' badge + greyed-out + disabled state on the
controls that don't have runtime consumers yet: Trigger Mode,
Autofix Mode, Autofix Severity Threshold, and Automatically fix CI
failures. SettingsRow grew a comingSoon prop to keep this consistent.
- My Settings drops the noop PR Preferences section and adds a Sign
Out button. preferred_pr_destination is removed from the profile
schema; old records get the field popped on next write.
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
2026-05-21 09:17:07 -07:00
|
|
|
from .enabled_repos import (
|
|
|
|
|
list_enabled_review_repos,
|
|
|
|
|
set_review_repo_enabled,
|
|
|
|
|
)
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
from .oauth import (
|
|
|
|
|
COOKIE_NAME,
|
|
|
|
|
SESSION_TTL_SECONDS,
|
|
|
|
|
STATE_COOKIE_NAME,
|
|
|
|
|
STATE_TTL_SECONDS,
|
|
|
|
|
decode_state,
|
|
|
|
|
exchange_code,
|
|
|
|
|
fetch_github_user,
|
|
|
|
|
hash_state_nonce,
|
|
|
|
|
issue_session,
|
|
|
|
|
issue_state,
|
|
|
|
|
new_state_nonce,
|
|
|
|
|
require_session,
|
|
|
|
|
sanitize_redirect_to,
|
|
|
|
|
)
|
|
|
|
|
from .options import SUPPORTED_MODELS
|
|
|
|
|
from .profiles import (
|
|
|
|
|
ProfileUpdate,
|
|
|
|
|
get_profile,
|
2026-05-21 10:44:04 -07:00
|
|
|
get_valid_access_token,
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
list_profiles,
|
2026-05-21 10:44:04 -07:00
|
|
|
upsert_access_token_from_github_response,
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
upsert_profile,
|
|
|
|
|
)
|
2026-05-21 10:44:04 -07:00
|
|
|
from .review_style_jobs import (
|
|
|
|
|
cancel_review_style_analysis,
|
|
|
|
|
start_review_style_analysis,
|
|
|
|
|
sync_review_style_run_status,
|
|
|
|
|
)
|
feat: tune reviewer for precision — web/wiki tools + recalibrated prompt (#1312)
* feat: tune reviewer for precision — web/wiki tools + recalibrated prompt
Reviewer agent now has web_search, fetch_url, and http_request alongside the
finding tools, so it can verify library semantics and consult the DeepWiki
auto-generated wiki for public repos (https://deepwiki.com/<owner>/<repo>)
before flagging cross-file or architectural concerns.
Prompt rewritten to push precision over recall:
- explicit severity ladder pushing reviews toward bimodal high/low instead of
defaulting to medium
- ≤200-char description target (gold set averages ~186 chars; we were at ~436)
- mandatory docs / wiki / code lookup before flagging concurrency, security,
or perf — the three categories that dominated false positives
- "do not flag" list covering compiler/linter-catchable nits, speculative
claims without a concrete attacker/interleaving/scale, style preferences
the codebase doesn't share, and test-quality nits on non-test diffs
- smart file-selection guidance for large PRs (deprioritize generated /
vendored / pure-rename hunks)
Eval config switched to openai:gpt-5.5 + high reasoning effort for the next
benchmark run.
* trim prompt
* subagent prompting
* confidence ratings
* added medium
* enforce confidence threshold
* .
* reviewer: precision-tuned prompt + drop confidence gate
Rewrites the reviewer system prompt around a defensibility bar (anchor +
failure mode + maintainer wouldn't say "not a bug"), an explicit do-not-file
list (style nits, speculation, scope-policing, same-bug fan-out), and a
checklist of 10 bug archetypes drawn from a per-PR audit of the eval golden
set. The audit showed 145 FPs in the last eval split ~28% speculative, ~26%
style-nit, ~31% real-but-unscored (mostly same-archetype fan-out); the new
prompt targets each class directly.
Confidence is still recorded on every finding for post-hoc calibration but
no longer gates publication — the audit showed the gate was a no-op (agent
self-rated 65% of findings "high" regardless), and the prompt's defensibility
bar is the actual discipline. Drops CONFIDENCE_ORDER, CONFIDENCE_THRESHOLD,
the confidence_threshold kwarg on filter_findings_for_publish, the
confidence_filtered score_mode, and the min_confidence kwarg on the eval
target's _extract_comments — all dead once the gate is gone.
Also removes the "informational" severity tier from the Severity enum,
SEVERITY_ORDER, and all validators / tests / docstrings. It was reserved for
FYI observations the dataset never rewards.
* benchmax
* adding google provider
* slight steering
* tuning
* more tuning
* fix
* cleanup
* reducing overfitting
* Add per-repo review style profiles and inject them into the reviewer.
Dashboard users can analyze historical PR review feedback per repository,
edit the resulting style guide, and have it loaded from LangGraph Store at
reviewer runtime (including Martian eval runs) keyed by owner/name.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix review style job errors leaking exception details to clients.
Return generic dashboard messages while logging full stack traces server-side.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-20 11:35:00 -07:00
|
|
|
from .review_styles import (
|
|
|
|
|
ReviewStyleCreate,
|
|
|
|
|
ReviewStylePromptUpdate,
|
|
|
|
|
create_review_style,
|
2026-05-21 10:44:04 -07:00
|
|
|
delete_review_style,
|
feat: tune reviewer for precision — web/wiki tools + recalibrated prompt (#1312)
* feat: tune reviewer for precision — web/wiki tools + recalibrated prompt
Reviewer agent now has web_search, fetch_url, and http_request alongside the
finding tools, so it can verify library semantics and consult the DeepWiki
auto-generated wiki for public repos (https://deepwiki.com/<owner>/<repo>)
before flagging cross-file or architectural concerns.
Prompt rewritten to push precision over recall:
- explicit severity ladder pushing reviews toward bimodal high/low instead of
defaulting to medium
- ≤200-char description target (gold set averages ~186 chars; we were at ~436)
- mandatory docs / wiki / code lookup before flagging concurrency, security,
or perf — the three categories that dominated false positives
- "do not flag" list covering compiler/linter-catchable nits, speculative
claims without a concrete attacker/interleaving/scale, style preferences
the codebase doesn't share, and test-quality nits on non-test diffs
- smart file-selection guidance for large PRs (deprioritize generated /
vendored / pure-rename hunks)
Eval config switched to openai:gpt-5.5 + high reasoning effort for the next
benchmark run.
* trim prompt
* subagent prompting
* confidence ratings
* added medium
* enforce confidence threshold
* .
* reviewer: precision-tuned prompt + drop confidence gate
Rewrites the reviewer system prompt around a defensibility bar (anchor +
failure mode + maintainer wouldn't say "not a bug"), an explicit do-not-file
list (style nits, speculation, scope-policing, same-bug fan-out), and a
checklist of 10 bug archetypes drawn from a per-PR audit of the eval golden
set. The audit showed 145 FPs in the last eval split ~28% speculative, ~26%
style-nit, ~31% real-but-unscored (mostly same-archetype fan-out); the new
prompt targets each class directly.
Confidence is still recorded on every finding for post-hoc calibration but
no longer gates publication — the audit showed the gate was a no-op (agent
self-rated 65% of findings "high" regardless), and the prompt's defensibility
bar is the actual discipline. Drops CONFIDENCE_ORDER, CONFIDENCE_THRESHOLD,
the confidence_threshold kwarg on filter_findings_for_publish, the
confidence_filtered score_mode, and the min_confidence kwarg on the eval
target's _extract_comments — all dead once the gate is gone.
Also removes the "informational" severity tier from the Severity enum,
SEVERITY_ORDER, and all validators / tests / docstrings. It was reserved for
FYI observations the dataset never rewards.
* benchmax
* adding google provider
* slight steering
* tuning
* more tuning
* fix
* cleanup
* reducing overfitting
* Add per-repo review style profiles and inject them into the reviewer.
Dashboard users can analyze historical PR review feedback per repository,
edit the resulting style guide, and have it loaded from LangGraph Store at
reviewer runtime (including Martian eval runs) keyed by owner/name.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix review style job errors leaking exception details to clients.
Return generic dashboard messages while logging full stack traces server-side.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-20 11:35:00 -07:00
|
|
|
get_review_style,
|
|
|
|
|
list_review_styles,
|
|
|
|
|
normalize_repo_full_name,
|
|
|
|
|
set_custom_prompt,
|
|
|
|
|
)
|
feat: refactor UI into sidebar layout (#1318)
* feat(dashboard): refactor UI into Cursor-style sidebar layout
Replace the top-bar AppHeader with a left sidebar that holds the four
primary sections (My Settings, Cloud Agents, Open SWE Review,
Integrations) and the user account menu at the bottom. Admin remains a
hidden route reachable from the sidebar only when is_admin.
UI:
- AppShell + AppSidebar with bottom-anchored avatar/sign-out menu.
- My Settings: profile (first/last name) + PR destination preference.
- Cloud Agents: model/effort, default repo, base branch, branch prefix,
PR toggles (auto-fix CI, create PRs, allow artifacts), Slack
notifications.
- Open SWE Review: trigger mode, draft reviews, PR summaries, autofix
mode + severity threshold, plus the existing per-repo review-style
prompts (was /review-styles).
- Integrations: GitHub install status + Slack/Linear status rows.
- New Switch and Menu primitives.
Backend:
- Extend ProfileUpdate with first_name, last_name, base_branch,
branch_prefix, auto_fix_ci, create_prs, allow_artifacts,
slack_notifications, preferred_pr_destination.
- Add team_settings module + /team-settings GET (any session) / PUT
(admin only) for the reviewer configuration.
* fix(dashboard): replace base-ui Menu with plain dropdown, stabilise Selects
- base-ui's Menu hit an "Invalid hook call / Cannot read properties of
null (reading 'useRef')" crash inside fastComponent under Vite's dep
optimisation. The sidebar account menu is the only consumer, so swap
it for a useState-driven dropdown with click-outside + Esc handling
and drop the unused Menu wrapper.
- Initialise the Open SWE Review settings form with the same defaults
the server returns instead of `null`, so the Selects don't switch
from uncontrolled (`undefined`) to controlled on first data load.
* fix(dashboard): address review feedback on admin overwrite + form reset
- admin PUT /admin/profiles/{login} only sent model/effort/repo from the
admin form. ProfileUpdate's Pydantic defaults then wrote auto_fix_ci /
create_prs / slack_notifications etc back onto the target user's
profile, clobbering whatever they had configured. Switch to
model_dump(exclude_unset=True) and overlay the incoming fields on the
user's existing stored profile so only sent fields change.
- my-settings.tsx and cloud-agents.tsx initialised local state from
profile.data on every change. Because useSaveProfile updates the
cached profile on success, toggling any control would overwrite the
user's typed-but-unsaved input in another field. Guard the
initialisation effect with a ref so it runs once on first load.
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
2026-05-20 14:13:23 -07:00
|
|
|
from .team_settings import (
|
|
|
|
|
TeamSettingsUpdate,
|
|
|
|
|
get_team_settings,
|
|
|
|
|
upsert_team_settings,
|
|
|
|
|
)
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
|
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
|
|
|
|
router = APIRouter(prefix="/dashboard/api", tags=["dashboard"])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _require_admin(session: dict[str, Any]) -> dict[str, Any]:
|
|
|
|
|
if not is_admin(session.get("email")):
|
|
|
|
|
raise HTTPException(403, "admin only")
|
|
|
|
|
return session
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
_SESSION_DEP = Depends(require_session)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _admin_session(session: dict[str, Any] = _SESSION_DEP) -> dict[str, Any]:
|
|
|
|
|
return _require_admin(session)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
_ADMIN_DEP = Depends(_admin_session)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _api_base_url() -> str:
|
|
|
|
|
v = os.environ.get("DASHBOARD_API_BASE_URL", "").rstrip("/")
|
|
|
|
|
if not v:
|
|
|
|
|
raise HTTPException(500, "DASHBOARD_API_BASE_URL not configured")
|
|
|
|
|
return v
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _frontend_base_url() -> str:
|
|
|
|
|
v = os.environ.get("DASHBOARD_BASE_URL", "").rstrip("/")
|
|
|
|
|
if not v:
|
|
|
|
|
raise HTTPException(500, "DASHBOARD_BASE_URL not configured")
|
|
|
|
|
return v
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _set_session_cookie(response: Response, jwt_token: str) -> None:
|
|
|
|
|
response.set_cookie(
|
|
|
|
|
key=COOKIE_NAME,
|
|
|
|
|
value=jwt_token,
|
|
|
|
|
max_age=SESSION_TTL_SECONDS,
|
|
|
|
|
httponly=True,
|
|
|
|
|
secure=True,
|
|
|
|
|
samesite="none",
|
|
|
|
|
path="/",
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _set_state_cookie(response: Response, nonce: str) -> None:
|
|
|
|
|
# SameSite=Lax so GitHub's top-level redirect back to /auth/callback
|
|
|
|
|
# still presents this cookie; the cookie is single-purpose and lives
|
|
|
|
|
# only for the duration of one OAuth round-trip.
|
|
|
|
|
response.set_cookie(
|
|
|
|
|
key=STATE_COOKIE_NAME,
|
|
|
|
|
value=nonce,
|
|
|
|
|
max_age=STATE_TTL_SECONDS,
|
|
|
|
|
httponly=True,
|
|
|
|
|
secure=True,
|
|
|
|
|
samesite="lax",
|
|
|
|
|
path="/dashboard/api/auth",
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _clear_state_cookie(response: Response) -> None:
|
|
|
|
|
response.delete_cookie(
|
|
|
|
|
STATE_COOKIE_NAME, path="/dashboard/api/auth", samesite="lax", secure=True
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.get("/auth/login")
|
|
|
|
|
async def auth_login(request: Request, redirect_to: str | None = None) -> RedirectResponse:
|
|
|
|
|
client_id = os.environ.get("GITHUB_APP_CLIENT_ID", "")
|
|
|
|
|
if not client_id:
|
|
|
|
|
raise HTTPException(500, "GITHUB_APP_CLIENT_ID not configured")
|
|
|
|
|
safe_redirect = sanitize_redirect_to(redirect_to) or _frontend_base_url()
|
|
|
|
|
|
|
|
|
|
nonce = new_state_nonce()
|
|
|
|
|
state = issue_state(redirect_to=safe_redirect, nonce_hash=hash_state_nonce(nonce))
|
|
|
|
|
redirect_uri = f"{_api_base_url()}/dashboard/api/auth/callback"
|
|
|
|
|
url = (
|
|
|
|
|
"https://github.com/login/oauth/authorize"
|
|
|
|
|
f"?client_id={client_id}"
|
|
|
|
|
f"&redirect_uri={redirect_uri}"
|
|
|
|
|
f"&state={state}"
|
|
|
|
|
)
|
|
|
|
|
response = RedirectResponse(url, status_code=302)
|
|
|
|
|
_set_state_cookie(response, nonce)
|
|
|
|
|
return response
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.get("/auth/callback")
|
|
|
|
|
async def auth_callback(request: Request, code: str, state: str) -> RedirectResponse:
|
|
|
|
|
state_payload = decode_state(state)
|
|
|
|
|
state_nonce_hash = state_payload.get("nonce_hash")
|
|
|
|
|
cookie_nonce = request.cookies.get(STATE_COOKIE_NAME)
|
|
|
|
|
if (
|
|
|
|
|
not isinstance(state_nonce_hash, str)
|
|
|
|
|
or not cookie_nonce
|
|
|
|
|
or not hmac.compare_digest(hash_state_nonce(cookie_nonce), state_nonce_hash)
|
|
|
|
|
):
|
|
|
|
|
# Either the cookie went missing (different browser, expired,
|
|
|
|
|
# cookies blocked) or the state was issued for a different session.
|
|
|
|
|
raise HTTPException(400, "oauth state mismatch — please retry login")
|
|
|
|
|
|
|
|
|
|
redirect_to = sanitize_redirect_to(state_payload.get("redirect_to")) or _frontend_base_url()
|
|
|
|
|
|
2026-05-21 10:44:04 -07:00
|
|
|
token_data = await exchange_code(code)
|
|
|
|
|
access_token = token_data.get("access_token")
|
|
|
|
|
if not isinstance(access_token, str):
|
|
|
|
|
raise HTTPException(400, "oauth exchange missing access_token")
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
user, email = await fetch_github_user(access_token)
|
|
|
|
|
login = user.get("login")
|
|
|
|
|
if not login:
|
|
|
|
|
raise HTTPException(400, "could not resolve GitHub login")
|
|
|
|
|
|
2026-05-21 10:44:04 -07:00
|
|
|
await upsert_access_token_from_github_response(login, email or "", token_data)
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
|
|
|
|
|
session_jwt = issue_session(login=login, email=email, avatar_url=user.get("avatar_url"))
|
|
|
|
|
response = RedirectResponse(redirect_to, status_code=302)
|
|
|
|
|
_set_session_cookie(response, session_jwt)
|
|
|
|
|
_clear_state_cookie(response)
|
|
|
|
|
return response
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.post("/auth/logout")
|
|
|
|
|
async def auth_logout() -> Response:
|
|
|
|
|
response = Response(status_code=204)
|
|
|
|
|
response.delete_cookie(COOKIE_NAME, path="/", samesite="none", secure=True)
|
|
|
|
|
return response
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.get("/me")
|
|
|
|
|
async def me(session: dict[str, Any] = _SESSION_DEP) -> dict[str, Any]:
|
|
|
|
|
return {
|
|
|
|
|
"login": session["sub"],
|
|
|
|
|
"email": session.get("email"),
|
|
|
|
|
"avatar_url": session.get("avatar_url"),
|
|
|
|
|
"is_admin": is_admin(session.get("email")),
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.get("/options")
|
|
|
|
|
async def options() -> dict[str, Any]:
|
|
|
|
|
return {"models": SUPPORTED_MODELS}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.get("/profile")
|
|
|
|
|
async def get_my_profile(
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
|
|
|
|
profile = await get_profile(session["sub"])
|
|
|
|
|
return profile or {}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.put("/profile")
|
|
|
|
|
async def put_my_profile(
|
|
|
|
|
update: ProfileUpdate,
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
|
|
|
|
update.validate_pairing()
|
|
|
|
|
return await upsert_profile(session["sub"], session.get("email") or "", update)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.get("/admin/profiles")
|
|
|
|
|
async def admin_list_profiles(
|
|
|
|
|
_admin: dict[str, Any] = _ADMIN_DEP,
|
|
|
|
|
) -> list[dict[str, Any]]:
|
|
|
|
|
return await list_profiles()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class AdminProfileUpdate(ProfileUpdate):
|
|
|
|
|
email: str | None = None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.put("/admin/profiles/{login}")
|
|
|
|
|
async def admin_put_profile(
|
|
|
|
|
login: str,
|
|
|
|
|
update: AdminProfileUpdate,
|
|
|
|
|
_admin: dict[str, Any] = _ADMIN_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
|
|
|
|
update.validate_pairing()
|
|
|
|
|
existing = await get_profile(login) or {}
|
|
|
|
|
email = update.email or existing.get("email") or ""
|
feat: refactor UI into sidebar layout (#1318)
* feat(dashboard): refactor UI into Cursor-style sidebar layout
Replace the top-bar AppHeader with a left sidebar that holds the four
primary sections (My Settings, Cloud Agents, Open SWE Review,
Integrations) and the user account menu at the bottom. Admin remains a
hidden route reachable from the sidebar only when is_admin.
UI:
- AppShell + AppSidebar with bottom-anchored avatar/sign-out menu.
- My Settings: profile (first/last name) + PR destination preference.
- Cloud Agents: model/effort, default repo, base branch, branch prefix,
PR toggles (auto-fix CI, create PRs, allow artifacts), Slack
notifications.
- Open SWE Review: trigger mode, draft reviews, PR summaries, autofix
mode + severity threshold, plus the existing per-repo review-style
prompts (was /review-styles).
- Integrations: GitHub install status + Slack/Linear status rows.
- New Switch and Menu primitives.
Backend:
- Extend ProfileUpdate with first_name, last_name, base_branch,
branch_prefix, auto_fix_ci, create_prs, allow_artifacts,
slack_notifications, preferred_pr_destination.
- Add team_settings module + /team-settings GET (any session) / PUT
(admin only) for the reviewer configuration.
* fix(dashboard): replace base-ui Menu with plain dropdown, stabilise Selects
- base-ui's Menu hit an "Invalid hook call / Cannot read properties of
null (reading 'useRef')" crash inside fastComponent under Vite's dep
optimisation. The sidebar account menu is the only consumer, so swap
it for a useState-driven dropdown with click-outside + Esc handling
and drop the unused Menu wrapper.
- Initialise the Open SWE Review settings form with the same defaults
the server returns instead of `null`, so the Selects don't switch
from uncontrolled (`undefined`) to controlled on first data load.
* fix(dashboard): address review feedback on admin overwrite + form reset
- admin PUT /admin/profiles/{login} only sent model/effort/repo from the
admin form. ProfileUpdate's Pydantic defaults then wrote auto_fix_ci /
create_prs / slack_notifications etc back onto the target user's
profile, clobbering whatever they had configured. Switch to
model_dump(exclude_unset=True) and overlay the incoming fields on the
user's existing stored profile so only sent fields change.
- my-settings.tsx and cloud-agents.tsx initialised local state from
profile.data on every change. Because useSaveProfile updates the
cached profile on success, toggling any control would overwrite the
user's typed-but-unsaved input in another field. Guard the
initialisation effect with a ref so it runs once on first load.
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
2026-05-20 14:13:23 -07:00
|
|
|
# Overlay only fields that were explicitly sent so the admin form (which
|
|
|
|
|
# only sends model/effort/repo) can't reset other fields the target user
|
|
|
|
|
# configured via My Settings / Cloud Agents to ProfileUpdate's defaults.
|
|
|
|
|
incoming = update.model_dump(exclude={"email"}, exclude_unset=True)
|
|
|
|
|
merged = {**existing, **incoming}
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
base = ProfileUpdate(
|
feat: refactor UI into sidebar layout (#1318)
* feat(dashboard): refactor UI into Cursor-style sidebar layout
Replace the top-bar AppHeader with a left sidebar that holds the four
primary sections (My Settings, Cloud Agents, Open SWE Review,
Integrations) and the user account menu at the bottom. Admin remains a
hidden route reachable from the sidebar only when is_admin.
UI:
- AppShell + AppSidebar with bottom-anchored avatar/sign-out menu.
- My Settings: profile (first/last name) + PR destination preference.
- Cloud Agents: model/effort, default repo, base branch, branch prefix,
PR toggles (auto-fix CI, create PRs, allow artifacts), Slack
notifications.
- Open SWE Review: trigger mode, draft reviews, PR summaries, autofix
mode + severity threshold, plus the existing per-repo review-style
prompts (was /review-styles).
- Integrations: GitHub install status + Slack/Linear status rows.
- New Switch and Menu primitives.
Backend:
- Extend ProfileUpdate with first_name, last_name, base_branch,
branch_prefix, auto_fix_ci, create_prs, allow_artifacts,
slack_notifications, preferred_pr_destination.
- Add team_settings module + /team-settings GET (any session) / PUT
(admin only) for the reviewer configuration.
* fix(dashboard): replace base-ui Menu with plain dropdown, stabilise Selects
- base-ui's Menu hit an "Invalid hook call / Cannot read properties of
null (reading 'useRef')" crash inside fastComponent under Vite's dep
optimisation. The sidebar account menu is the only consumer, so swap
it for a useState-driven dropdown with click-outside + Esc handling
and drop the unused Menu wrapper.
- Initialise the Open SWE Review settings form with the same defaults
the server returns instead of `null`, so the Selects don't switch
from uncontrolled (`undefined`) to controlled on first data load.
* fix(dashboard): address review feedback on admin overwrite + form reset
- admin PUT /admin/profiles/{login} only sent model/effort/repo from the
admin form. ProfileUpdate's Pydantic defaults then wrote auto_fix_ci /
create_prs / slack_notifications etc back onto the target user's
profile, clobbering whatever they had configured. Switch to
model_dump(exclude_unset=True) and overlay the incoming fields on the
user's existing stored profile so only sent fields change.
- my-settings.tsx and cloud-agents.tsx initialised local state from
profile.data on every change. Because useSaveProfile updates the
cached profile on success, toggling any control would overwrite the
user's typed-but-unsaved input in another field. Guard the
initialisation effect with a ref so it runs once on first load.
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
2026-05-20 14:13:23 -07:00
|
|
|
**{k: v for k, v in merged.items() if k in ProfileUpdate.model_fields},
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
)
|
|
|
|
|
return await upsert_profile(login, email, base)
|
|
|
|
|
|
|
|
|
|
|
feat: refactor UI into sidebar layout (#1318)
* feat(dashboard): refactor UI into Cursor-style sidebar layout
Replace the top-bar AppHeader with a left sidebar that holds the four
primary sections (My Settings, Cloud Agents, Open SWE Review,
Integrations) and the user account menu at the bottom. Admin remains a
hidden route reachable from the sidebar only when is_admin.
UI:
- AppShell + AppSidebar with bottom-anchored avatar/sign-out menu.
- My Settings: profile (first/last name) + PR destination preference.
- Cloud Agents: model/effort, default repo, base branch, branch prefix,
PR toggles (auto-fix CI, create PRs, allow artifacts), Slack
notifications.
- Open SWE Review: trigger mode, draft reviews, PR summaries, autofix
mode + severity threshold, plus the existing per-repo review-style
prompts (was /review-styles).
- Integrations: GitHub install status + Slack/Linear status rows.
- New Switch and Menu primitives.
Backend:
- Extend ProfileUpdate with first_name, last_name, base_branch,
branch_prefix, auto_fix_ci, create_prs, allow_artifacts,
slack_notifications, preferred_pr_destination.
- Add team_settings module + /team-settings GET (any session) / PUT
(admin only) for the reviewer configuration.
* fix(dashboard): replace base-ui Menu with plain dropdown, stabilise Selects
- base-ui's Menu hit an "Invalid hook call / Cannot read properties of
null (reading 'useRef')" crash inside fastComponent under Vite's dep
optimisation. The sidebar account menu is the only consumer, so swap
it for a useState-driven dropdown with click-outside + Esc handling
and drop the unused Menu wrapper.
- Initialise the Open SWE Review settings form with the same defaults
the server returns instead of `null`, so the Selects don't switch
from uncontrolled (`undefined`) to controlled on first data load.
* fix(dashboard): address review feedback on admin overwrite + form reset
- admin PUT /admin/profiles/{login} only sent model/effort/repo from the
admin form. ProfileUpdate's Pydantic defaults then wrote auto_fix_ci /
create_prs / slack_notifications etc back onto the target user's
profile, clobbering whatever they had configured. Switch to
model_dump(exclude_unset=True) and overlay the incoming fields on the
user's existing stored profile so only sent fields change.
- my-settings.tsx and cloud-agents.tsx initialised local state from
profile.data on every change. Because useSaveProfile updates the
cached profile on success, toggling any control would overwrite the
user's typed-but-unsaved input in another field. Guard the
initialisation effect with a ref so it runs once on first load.
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
2026-05-20 14:13:23 -07:00
|
|
|
@router.get("/team-settings")
|
|
|
|
|
async def api_get_team_settings(
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
|
|
|
|
return await get_team_settings()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.put("/team-settings")
|
|
|
|
|
async def api_put_team_settings(
|
|
|
|
|
update: TeamSettingsUpdate,
|
|
|
|
|
_admin: dict[str, Any] = _ADMIN_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
|
|
|
|
return await upsert_team_settings(update)
|
|
|
|
|
|
|
|
|
|
|
feat: restructure Open SWE Review tab + wire create_prs (#1319)
* feat(dashboard): restructure Open SWE Review tab + wire create_prs
Restructures the dashboard around two related changes the reviewer settings
have been asking for:
- Wire profile.create_prs. Defaults to true (opt-out); when off the system
prompt gets a `Pull Request Policy Override` section telling the agent
to push the branch and notify with the branch URL instead of opening a
PR. Removes the noop Slack Notifications / Allow Artifacts / First Name
/ Last Name controls and their schema fields.
- Repositories opt-in for Open SWE Review. New per-team enabled list
stored in the LangGraph Store (`["enabled_review_repos"]`). Every
reviewer webhook chokepoint now goes through `_is_repo_enabled_for_review`
which AND-combines the existing env allowlist with the dashboard list.
Default is empty (opt-in) — admins enable repos per-installation from
the new Repositories page nested under Open SWE Review.
- Open SWE Review tab now mirrors the Cursor "rules" pattern: main page
shows installation rows + a Rules entry; both drill into nested pages
(/review/repositories/$owner and /review/styles) with a back link.
- Adds the new logo/favicon assets shipped from sidebar + html head.
Tests pass with a new autouse fixture (`tests/conftest.py`) that defaults
`is_review_repo_enabled` to True for existing allowlist tests.
* fix(dashboard): make main content scroll independently of the sidebar
Outer flex container was min-h-svh, so it grew with main's content and the
whole page scrolled — sidebar moved with it. Pin to h-svh + overflow-hidden
so the sidebar stays put and only <main> scrolls.
* fix(dashboard): make disabled repo toggles obviously disabled
Switch's disabled state used opacity-50 against a muted background, so
the not-admin state looked nearly identical to the off state. Bump to
opacity-40 + grayscale, and wrap each repo toggle in a span carrying a
native hover tooltip explaining why it's disabled.
* fix(switch): handle base-ui's data-disabled state
base-ui's Switch.Root sets data-disabled (not the HTML disabled attribute)
when disabled, so Tailwind's disabled: variant never matches and the
button keeps its cursor-pointer + clickable look. Mirror the styling
under the data-[disabled] variant and add pointer-events-none so the
disabled state is both visible and actually unclickable.
* feat(dashboard): paginate per-installation repository list
20 repos per page with Prev / page X of Y / Next controls at the bottom.
Pager only renders when there are more than 20 repos. Page resets to 0
when navigating between installations.
* feat(dashboard): global default model selectors for Agent + Reviewer
Adds team-wide default model + reasoning effort for both agents in the
Admin tab so operators can switch models without redeploying.
Resolution chain:
Agent: hardcoded -> LLM_MODEL_ID env -> team default -> user profile
Reviewer: hardcoded -> LLM_MODEL_ID env -> team default -> per-call configurable
Team defaults live in team_settings and are validated against the
SUPPORTED_MODELS allowlist + the model's supported reasoning efforts.
'Inherit from env' clears the override and falls back to LLM_MODEL_ID.
* refactor(models): drop LLM_MODEL_ID env in favour of the team default
The team default is now the single source of truth for the runtime model
choice; per-user (agent) and per-call configurable (reviewer) selections
still win on top. When no admin has touched the team default, it surfaces
the hardcoded fallback (DEFAULT_MODEL_ID + its default effort), so the
admin UI's dropdown is always pre-populated with a sensible value.
The Admin UI loses the 'Inherit from env' option since there is no longer
an env layer to inherit from.
* chore(models): set hardcoded fallback to gpt-5.5 medium
Decouple the team-default boot value (gpt-5.5 / medium) from each model's
ProfileForm-suggested default_effort so we can change one without nudging
the other. The Opus xhigh default for new user profiles is unchanged.
* feat(dashboard): trigger-mode copy, Coming Soon badges, logout in My Settings
- Rename trigger mode 'ready_for_review' -> 'once_per_pr' with new
description copy that matches the screenshot. Legacy stored values
fall back to 'every_push' on read so the UI never shows an unknown
selection.
- Add a 'Coming soon' badge + greyed-out + disabled state on the
controls that don't have runtime consumers yet: Trigger Mode,
Autofix Mode, Autofix Severity Threshold, and Automatically fix CI
failures. SettingsRow grew a comingSoon prop to keep this consistent.
- My Settings drops the noop PR Preferences section and adds a Sign
Out button. preferred_pr_destination is removed from the profile
schema; old records get the field popped on next write.
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
2026-05-21 09:17:07 -07:00
|
|
|
class EnabledReviewRepoUpdate(BaseModel):
|
|
|
|
|
full_name: str
|
|
|
|
|
enabled: bool
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.get("/enabled-review-repos")
|
|
|
|
|
async def api_list_enabled_review_repos(
|
|
|
|
|
_session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> dict[str, list[str]]:
|
|
|
|
|
return {"repos": await list_enabled_review_repos()}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.put("/enabled-review-repos")
|
|
|
|
|
async def api_set_enabled_review_repo(
|
|
|
|
|
update: EnabledReviewRepoUpdate,
|
|
|
|
|
_admin: dict[str, Any] = _ADMIN_DEP,
|
|
|
|
|
) -> dict[str, list[str]]:
|
|
|
|
|
repos = await set_review_repo_enabled(update.full_name, update.enabled)
|
|
|
|
|
return {"repos": repos}
|
|
|
|
|
|
|
|
|
|
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
def _next_link_url(link_header: str | None) -> str | None:
|
|
|
|
|
if not link_header:
|
|
|
|
|
return None
|
|
|
|
|
# GitHub Link header is comma-separated: '<url>; rel="next", <url>; rel="last"'
|
|
|
|
|
for part in link_header.split(","):
|
|
|
|
|
segments = [s.strip() for s in part.split(";")]
|
|
|
|
|
if len(segments) >= 2 and 'rel="next"' in segments[1] and segments[0].startswith("<"):
|
|
|
|
|
return segments[0][1:-1]
|
|
|
|
|
return None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
async def _paginate(
|
|
|
|
|
client: httpx.AsyncClient,
|
|
|
|
|
url: str,
|
|
|
|
|
*,
|
|
|
|
|
headers: dict[str, str],
|
|
|
|
|
items_key: str | None,
|
|
|
|
|
cap: int = 1000,
|
|
|
|
|
) -> list[dict[str, Any]]:
|
|
|
|
|
"""Follow ``Link: rel="next"`` until exhausted (or cap reached).
|
|
|
|
|
|
|
|
|
|
``items_key`` is the JSON key holding the list when the endpoint returns
|
|
|
|
|
a wrapper object (e.g. ``/user/installations`` returns
|
|
|
|
|
``{"total_count": N, "installations": [...]}``). When ``None`` the
|
|
|
|
|
response body itself is treated as the list.
|
|
|
|
|
"""
|
|
|
|
|
out: list[dict[str, Any]] = []
|
|
|
|
|
next_url: str | None = url
|
|
|
|
|
first = True
|
|
|
|
|
while next_url and len(out) < cap:
|
|
|
|
|
params = {"per_page": "100"} if first else None
|
|
|
|
|
r = await client.get(next_url, headers=headers, params=params)
|
|
|
|
|
if r.status_code == 401:
|
|
|
|
|
raise HTTPException(401, "github token expired, re-login required")
|
|
|
|
|
r.raise_for_status()
|
|
|
|
|
body = r.json()
|
|
|
|
|
page = body.get(items_key, []) if items_key else body
|
|
|
|
|
if isinstance(page, list):
|
|
|
|
|
out.extend(page)
|
|
|
|
|
next_url = _next_link_url(r.headers.get("Link"))
|
|
|
|
|
first = False
|
|
|
|
|
return out
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.get("/repos")
|
|
|
|
|
async def list_repos(
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
|
|
|
|
"""List repos where open-swe is installed and the user has access.
|
|
|
|
|
|
|
|
|
|
Paginates both ``/user/installations`` and per-installation
|
|
|
|
|
``/user/installations/{id}/repositories`` so users with multiple
|
|
|
|
|
installations or >30 accessible repos get the complete set.
|
|
|
|
|
"""
|
2026-05-21 10:44:04 -07:00
|
|
|
login = session["sub"]
|
|
|
|
|
token = await get_valid_access_token(login)
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
if not token:
|
|
|
|
|
raise HTTPException(401, "github token unavailable, re-login required")
|
|
|
|
|
headers = {
|
|
|
|
|
"Authorization": f"Bearer {token}",
|
|
|
|
|
"Accept": "application/vnd.github+json",
|
|
|
|
|
"X-GitHub-Api-Version": "2022-11-28",
|
|
|
|
|
}
|
|
|
|
|
async with httpx.AsyncClient() as client:
|
2026-05-21 10:44:04 -07:00
|
|
|
try:
|
|
|
|
|
installations = await _paginate(
|
|
|
|
|
client,
|
|
|
|
|
"https://api.github.com/user/installations",
|
|
|
|
|
headers=headers,
|
|
|
|
|
items_key="installations",
|
|
|
|
|
)
|
|
|
|
|
except HTTPException as exc:
|
|
|
|
|
if exc.status_code != 401:
|
|
|
|
|
raise
|
|
|
|
|
token = await get_valid_access_token(login, force_refresh=True)
|
|
|
|
|
if not token:
|
|
|
|
|
raise HTTPException(401, "github token expired, re-login required") from exc
|
|
|
|
|
headers["Authorization"] = f"Bearer {token}"
|
|
|
|
|
installations = await _paginate(
|
|
|
|
|
client,
|
|
|
|
|
"https://api.github.com/user/installations",
|
|
|
|
|
headers=headers,
|
|
|
|
|
items_key="installations",
|
|
|
|
|
)
|
feat: open-swe dashboard for per-user profile config (#1302)
* feat: dashboard backend — GitHub OAuth, profile CRUD, admin endpoints
Adds agent/dashboard/ FastAPI router mounted at /dashboard/api covering:
- GitHub App OAuth login → JWT cookie session (cross-domain ready)
- profile CRUD against LangGraph Store with model+effort validation
- admin gate via CONFIGURED_ADMINS
- /repos via /user/installations using the user's encrypted OAuth token
CORS allowlist on webapp.py is opt-in via DASHBOARD_ALLOWED_ORIGINS so the
Vercel-hosted frontend can call the LangSmith deployment with credentials.
* feat: apply dashboard profile model/effort overrides in get_agent
Look up the triggering user's GitHub login from config (direct field or
GITHUB_USER_EMAIL_MAP reverse lookup), read their profile from the Store,
and apply default_model + reasoning_effort to make_model when both are
valid. Effort 'max' is captured on the profile but not yet wired through —
the OpenAI Reasoning Literal doesn't accept it.
* feat: ui/ TanStack Start dashboard for profile config
Scaffolded with the shadcn b7CScJIjA preset (TanStack Start template,
base-ui primitives, Tailwind v4). Three routes:
- /login — Sign in with GitHub (links to /dashboard/api/auth/login)
- /profile — Edit default model, reasoning effort, default repo
- /admin — Admin-only: list users and edit other profiles
API client (src/lib/api.ts) uses credentials: include so the osw_session
cookie set by the OAuth callback rides cross-origin. VITE_DASHBOARD_API_BASE_URL
points at the LangSmith deployment.
Effort options re-render when the model changes; 'max' on Opus 4.7 is
captured on the profile but ignored downstream until anthropic reasoning
is wired through make_model.
* feat: searchable Combobox for default repo picker
Replaces the Select with a base-ui Combobox so users can filter by typing,
the popup is wider than the trigger so full owner/repo names are readable,
and the list caps at max-h-80 to stay on screen.
* fix: address review comments + wire default_repo and Anthropic thinking
Security/correctness fixes from PR review:
* Open redirect: validate `redirect_to` in `/auth/login` against
`DASHBOARD_BASE_URL` + `DASHBOARD_ALLOWED_ORIGINS` before signing it
into the state JWT. Anything off-allowlist falls back to the dashboard
base URL. (PR #1302 r3250054386)
* Login CSRF: bind the OAuth `state` to the requesting browser. At
`/auth/login` we generate a fresh nonce, set it as a short-lived
HttpOnly SameSite=Lax cookie scoped to `/dashboard/api/auth`, and
embed `hash_state_nonce(nonce)` in the state JWT. At `/auth/callback`
we require the cookie nonce to hash-match the state JWT's nonce_hash
(constant-time compare). (PR #1302 r3250054395)
* RMW race in profile vs token writes: split storage into two
namespaces — `["profiles"]` for user-editable settings and
`["oauth_tokens"]` for the encrypted GitHub token. Each upsert now
only writes its own namespace so an in-flight profile save can no
longer clobber a fresh token from a concurrent re-login (and vice
versa). (PR #1302 r3250054393)
* /repos pagination: follow `Link: rel="next"` for both
`/user/installations` and per-installation `/repositories` with
per_page=100, capped at 1000 items. (PR #1302 r3250054401)
Feature wires:
* default_repo: applied as a fallback in `get_slack_repo_config` (after
explicit-repo / thread metadata, before the env defaults) and in the
Linear webhook (after comment-body extraction, before team mapping).
Both paths resolve the triggering user's GitHub login via
GITHUB_USER_EMAIL_MAP and read the profile's default_repo.
* Anthropic "thinking" effort: `make_model` now accepts a `thinking`
kwarg; `get_agent` maps profile effort {low,medium,high,xhigh,max}
to budget_tokens {1k,4k,12k,32k,60k} when the chosen model is
anthropic. OpenAI path still ignores "max" since the Literal doesn't
accept it.
2026-05-15 11:23:53 -07:00
|
|
|
repositories: list[dict[str, Any]] = []
|
|
|
|
|
for inst in installations:
|
|
|
|
|
inst_id = inst.get("id")
|
|
|
|
|
if inst_id is None:
|
|
|
|
|
continue
|
|
|
|
|
try:
|
|
|
|
|
repos = await _paginate(
|
|
|
|
|
client,
|
|
|
|
|
f"https://api.github.com/user/installations/{inst_id}/repositories",
|
|
|
|
|
headers=headers,
|
|
|
|
|
items_key="repositories",
|
|
|
|
|
)
|
|
|
|
|
except HTTPException:
|
|
|
|
|
raise
|
|
|
|
|
except httpx.HTTPStatusError:
|
|
|
|
|
continue
|
|
|
|
|
repositories.extend(repos)
|
|
|
|
|
return {
|
|
|
|
|
"installations": [
|
|
|
|
|
{
|
|
|
|
|
"id": i.get("id"),
|
|
|
|
|
"account": (i.get("account") or {}).get("login"),
|
|
|
|
|
"account_type": (i.get("account") or {}).get("type"),
|
|
|
|
|
}
|
|
|
|
|
for i in installations
|
|
|
|
|
],
|
|
|
|
|
"repositories": [
|
|
|
|
|
{"full_name": r.get("full_name"), "private": r.get("private", False)}
|
|
|
|
|
for r in repositories
|
|
|
|
|
if r.get("full_name")
|
|
|
|
|
],
|
|
|
|
|
}
|
feat: tune reviewer for precision — web/wiki tools + recalibrated prompt (#1312)
* feat: tune reviewer for precision — web/wiki tools + recalibrated prompt
Reviewer agent now has web_search, fetch_url, and http_request alongside the
finding tools, so it can verify library semantics and consult the DeepWiki
auto-generated wiki for public repos (https://deepwiki.com/<owner>/<repo>)
before flagging cross-file or architectural concerns.
Prompt rewritten to push precision over recall:
- explicit severity ladder pushing reviews toward bimodal high/low instead of
defaulting to medium
- ≤200-char description target (gold set averages ~186 chars; we were at ~436)
- mandatory docs / wiki / code lookup before flagging concurrency, security,
or perf — the three categories that dominated false positives
- "do not flag" list covering compiler/linter-catchable nits, speculative
claims without a concrete attacker/interleaving/scale, style preferences
the codebase doesn't share, and test-quality nits on non-test diffs
- smart file-selection guidance for large PRs (deprioritize generated /
vendored / pure-rename hunks)
Eval config switched to openai:gpt-5.5 + high reasoning effort for the next
benchmark run.
* trim prompt
* subagent prompting
* confidence ratings
* added medium
* enforce confidence threshold
* .
* reviewer: precision-tuned prompt + drop confidence gate
Rewrites the reviewer system prompt around a defensibility bar (anchor +
failure mode + maintainer wouldn't say "not a bug"), an explicit do-not-file
list (style nits, speculation, scope-policing, same-bug fan-out), and a
checklist of 10 bug archetypes drawn from a per-PR audit of the eval golden
set. The audit showed 145 FPs in the last eval split ~28% speculative, ~26%
style-nit, ~31% real-but-unscored (mostly same-archetype fan-out); the new
prompt targets each class directly.
Confidence is still recorded on every finding for post-hoc calibration but
no longer gates publication — the audit showed the gate was a no-op (agent
self-rated 65% of findings "high" regardless), and the prompt's defensibility
bar is the actual discipline. Drops CONFIDENCE_ORDER, CONFIDENCE_THRESHOLD,
the confidence_threshold kwarg on filter_findings_for_publish, the
confidence_filtered score_mode, and the min_confidence kwarg on the eval
target's _extract_comments — all dead once the gate is gone.
Also removes the "informational" severity tier from the Severity enum,
SEVERITY_ORDER, and all validators / tests / docstrings. It was reserved for
FYI observations the dataset never rewards.
* benchmax
* adding google provider
* slight steering
* tuning
* more tuning
* fix
* cleanup
* reducing overfitting
* Add per-repo review style profiles and inject them into the reviewer.
Dashboard users can analyze historical PR review feedback per repository,
edit the resulting style guide, and have it loaded from LangGraph Store at
reviewer runtime (including Martian eval runs) keyed by owner/name.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix review style job errors leaking exception details to clients.
Return generic dashboard messages while logging full stack traces server-side.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-20 11:35:00 -07:00
|
|
|
|
|
|
|
|
|
2026-05-21 10:44:04 -07:00
|
|
|
def _raise_for_github_repo_status(status_code: int) -> None:
|
|
|
|
|
if status_code == 401:
|
|
|
|
|
raise HTTPException(401, "github token expired, re-login required")
|
|
|
|
|
if status_code == 404:
|
|
|
|
|
raise HTTPException(404, "repository not found")
|
|
|
|
|
if status_code == 403:
|
|
|
|
|
raise HTTPException(403, "no access to this private repository")
|
|
|
|
|
if status_code != 200:
|
|
|
|
|
raise HTTPException(502, f"github API error ({status_code})")
|
|
|
|
|
|
|
|
|
|
|
feat: tune reviewer for precision — web/wiki tools + recalibrated prompt (#1312)
* feat: tune reviewer for precision — web/wiki tools + recalibrated prompt
Reviewer agent now has web_search, fetch_url, and http_request alongside the
finding tools, so it can verify library semantics and consult the DeepWiki
auto-generated wiki for public repos (https://deepwiki.com/<owner>/<repo>)
before flagging cross-file or architectural concerns.
Prompt rewritten to push precision over recall:
- explicit severity ladder pushing reviews toward bimodal high/low instead of
defaulting to medium
- ≤200-char description target (gold set averages ~186 chars; we were at ~436)
- mandatory docs / wiki / code lookup before flagging concurrency, security,
or perf — the three categories that dominated false positives
- "do not flag" list covering compiler/linter-catchable nits, speculative
claims without a concrete attacker/interleaving/scale, style preferences
the codebase doesn't share, and test-quality nits on non-test diffs
- smart file-selection guidance for large PRs (deprioritize generated /
vendored / pure-rename hunks)
Eval config switched to openai:gpt-5.5 + high reasoning effort for the next
benchmark run.
* trim prompt
* subagent prompting
* confidence ratings
* added medium
* enforce confidence threshold
* .
* reviewer: precision-tuned prompt + drop confidence gate
Rewrites the reviewer system prompt around a defensibility bar (anchor +
failure mode + maintainer wouldn't say "not a bug"), an explicit do-not-file
list (style nits, speculation, scope-policing, same-bug fan-out), and a
checklist of 10 bug archetypes drawn from a per-PR audit of the eval golden
set. The audit showed 145 FPs in the last eval split ~28% speculative, ~26%
style-nit, ~31% real-but-unscored (mostly same-archetype fan-out); the new
prompt targets each class directly.
Confidence is still recorded on every finding for post-hoc calibration but
no longer gates publication — the audit showed the gate was a no-op (agent
self-rated 65% of findings "high" regardless), and the prompt's defensibility
bar is the actual discipline. Drops CONFIDENCE_ORDER, CONFIDENCE_THRESHOLD,
the confidence_threshold kwarg on filter_findings_for_publish, the
confidence_filtered score_mode, and the min_confidence kwarg on the eval
target's _extract_comments — all dead once the gate is gone.
Also removes the "informational" severity tier from the Severity enum,
SEVERITY_ORDER, and all validators / tests / docstrings. It was reserved for
FYI observations the dataset never rewards.
* benchmax
* adding google provider
* slight steering
* tuning
* more tuning
* fix
* cleanup
* reducing overfitting
* Add per-repo review style profiles and inject them into the reviewer.
Dashboard users can analyze historical PR review feedback per repository,
edit the resulting style guide, and have it loaded from LangGraph Store at
reviewer runtime (including Martian eval runs) keyed by owner/name.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix review style job errors leaking exception details to clients.
Return generic dashboard messages while logging full stack traces server-side.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-20 11:35:00 -07:00
|
|
|
async def _assert_repo_available_for_style_analysis(full_name: str, token: str) -> None:
|
|
|
|
|
"""Ensure the repo exists and is readable for style learning.
|
|
|
|
|
|
|
|
|
|
Public repositories are allowed without the GitHub App installed on them.
|
|
|
|
|
Private repositories require the authenticated user to have read access.
|
|
|
|
|
"""
|
|
|
|
|
full_name = normalize_repo_full_name(full_name)
|
|
|
|
|
headers = {
|
|
|
|
|
"Authorization": f"Bearer {token}",
|
|
|
|
|
"Accept": "application/vnd.github+json",
|
|
|
|
|
"X-GitHub-Api-Version": "2022-11-28",
|
|
|
|
|
}
|
|
|
|
|
owner, name = full_name.split("/", 1)
|
|
|
|
|
async with httpx.AsyncClient() as client:
|
|
|
|
|
r = await client.get(
|
|
|
|
|
f"https://api.github.com/repos/{owner}/{name}",
|
|
|
|
|
headers=headers,
|
|
|
|
|
)
|
2026-05-21 10:44:04 -07:00
|
|
|
_raise_for_github_repo_status(r.status_code)
|
feat: tune reviewer for precision — web/wiki tools + recalibrated prompt (#1312)
* feat: tune reviewer for precision — web/wiki tools + recalibrated prompt
Reviewer agent now has web_search, fetch_url, and http_request alongside the
finding tools, so it can verify library semantics and consult the DeepWiki
auto-generated wiki for public repos (https://deepwiki.com/<owner>/<repo>)
before flagging cross-file or architectural concerns.
Prompt rewritten to push precision over recall:
- explicit severity ladder pushing reviews toward bimodal high/low instead of
defaulting to medium
- ≤200-char description target (gold set averages ~186 chars; we were at ~436)
- mandatory docs / wiki / code lookup before flagging concurrency, security,
or perf — the three categories that dominated false positives
- "do not flag" list covering compiler/linter-catchable nits, speculative
claims without a concrete attacker/interleaving/scale, style preferences
the codebase doesn't share, and test-quality nits on non-test diffs
- smart file-selection guidance for large PRs (deprioritize generated /
vendored / pure-rename hunks)
Eval config switched to openai:gpt-5.5 + high reasoning effort for the next
benchmark run.
* trim prompt
* subagent prompting
* confidence ratings
* added medium
* enforce confidence threshold
* .
* reviewer: precision-tuned prompt + drop confidence gate
Rewrites the reviewer system prompt around a defensibility bar (anchor +
failure mode + maintainer wouldn't say "not a bug"), an explicit do-not-file
list (style nits, speculation, scope-policing, same-bug fan-out), and a
checklist of 10 bug archetypes drawn from a per-PR audit of the eval golden
set. The audit showed 145 FPs in the last eval split ~28% speculative, ~26%
style-nit, ~31% real-but-unscored (mostly same-archetype fan-out); the new
prompt targets each class directly.
Confidence is still recorded on every finding for post-hoc calibration but
no longer gates publication — the audit showed the gate was a no-op (agent
self-rated 65% of findings "high" regardless), and the prompt's defensibility
bar is the actual discipline. Drops CONFIDENCE_ORDER, CONFIDENCE_THRESHOLD,
the confidence_threshold kwarg on filter_findings_for_publish, the
confidence_filtered score_mode, and the min_confidence kwarg on the eval
target's _extract_comments — all dead once the gate is gone.
Also removes the "informational" severity tier from the Severity enum,
SEVERITY_ORDER, and all validators / tests / docstrings. It was reserved for
FYI observations the dataset never rewards.
* benchmax
* adding google provider
* slight steering
* tuning
* more tuning
* fix
* cleanup
* reducing overfitting
* Add per-repo review style profiles and inject them into the reviewer.
Dashboard users can analyze historical PR review feedback per repository,
edit the resulting style guide, and have it loaded from LangGraph Store at
reviewer runtime (including Martian eval runs) keyed by owner/name.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix review style job errors leaking exception details to clients.
Return generic dashboard messages while logging full stack traces server-side.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-20 11:35:00 -07:00
|
|
|
body = r.json()
|
|
|
|
|
if body.get("private") is not True:
|
|
|
|
|
return
|
|
|
|
|
# Private repo: 200 from GitHub implies the user's token can read it.
|
|
|
|
|
|
|
|
|
|
|
2026-05-21 10:44:04 -07:00
|
|
|
async def _require_repo_access_for_user(login: str, full_name: str) -> str:
|
|
|
|
|
"""Verify the user can read ``full_name`` on GitHub; return a valid access token."""
|
|
|
|
|
token = await get_valid_access_token(login)
|
|
|
|
|
if not token:
|
|
|
|
|
raise HTTPException(401, "github token unavailable, re-login required")
|
|
|
|
|
try:
|
|
|
|
|
await _assert_repo_available_for_style_analysis(full_name, token)
|
|
|
|
|
except HTTPException as exc:
|
|
|
|
|
if exc.status_code != 401:
|
|
|
|
|
raise
|
|
|
|
|
token = await get_valid_access_token(login, force_refresh=True)
|
|
|
|
|
if not token:
|
|
|
|
|
raise HTTPException(401, "github token expired, re-login required") from exc
|
|
|
|
|
await _assert_repo_available_for_style_analysis(full_name, token)
|
|
|
|
|
return token
|
|
|
|
|
|
|
|
|
|
|
feat: tune reviewer for precision — web/wiki tools + recalibrated prompt (#1312)
* feat: tune reviewer for precision — web/wiki tools + recalibrated prompt
Reviewer agent now has web_search, fetch_url, and http_request alongside the
finding tools, so it can verify library semantics and consult the DeepWiki
auto-generated wiki for public repos (https://deepwiki.com/<owner>/<repo>)
before flagging cross-file or architectural concerns.
Prompt rewritten to push precision over recall:
- explicit severity ladder pushing reviews toward bimodal high/low instead of
defaulting to medium
- ≤200-char description target (gold set averages ~186 chars; we were at ~436)
- mandatory docs / wiki / code lookup before flagging concurrency, security,
or perf — the three categories that dominated false positives
- "do not flag" list covering compiler/linter-catchable nits, speculative
claims without a concrete attacker/interleaving/scale, style preferences
the codebase doesn't share, and test-quality nits on non-test diffs
- smart file-selection guidance for large PRs (deprioritize generated /
vendored / pure-rename hunks)
Eval config switched to openai:gpt-5.5 + high reasoning effort for the next
benchmark run.
* trim prompt
* subagent prompting
* confidence ratings
* added medium
* enforce confidence threshold
* .
* reviewer: precision-tuned prompt + drop confidence gate
Rewrites the reviewer system prompt around a defensibility bar (anchor +
failure mode + maintainer wouldn't say "not a bug"), an explicit do-not-file
list (style nits, speculation, scope-policing, same-bug fan-out), and a
checklist of 10 bug archetypes drawn from a per-PR audit of the eval golden
set. The audit showed 145 FPs in the last eval split ~28% speculative, ~26%
style-nit, ~31% real-but-unscored (mostly same-archetype fan-out); the new
prompt targets each class directly.
Confidence is still recorded on every finding for post-hoc calibration but
no longer gates publication — the audit showed the gate was a no-op (agent
self-rated 65% of findings "high" regardless), and the prompt's defensibility
bar is the actual discipline. Drops CONFIDENCE_ORDER, CONFIDENCE_THRESHOLD,
the confidence_threshold kwarg on filter_findings_for_publish, the
confidence_filtered score_mode, and the min_confidence kwarg on the eval
target's _extract_comments — all dead once the gate is gone.
Also removes the "informational" severity tier from the Severity enum,
SEVERITY_ORDER, and all validators / tests / docstrings. It was reserved for
FYI observations the dataset never rewards.
* benchmax
* adding google provider
* slight steering
* tuning
* more tuning
* fix
* cleanup
* reducing overfitting
* Add per-repo review style profiles and inject them into the reviewer.
Dashboard users can analyze historical PR review feedback per repository,
edit the resulting style guide, and have it loaded from LangGraph Store at
reviewer runtime (including Martian eval runs) keyed by owner/name.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix review style job errors leaking exception details to clients.
Return generic dashboard messages while logging full stack traces server-side.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-20 11:35:00 -07:00
|
|
|
@router.get("/review-styles")
|
|
|
|
|
async def api_list_review_styles(
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> list[dict[str, Any]]:
|
|
|
|
|
records = await list_review_styles()
|
|
|
|
|
out: list[dict[str, Any]] = []
|
|
|
|
|
for record in records:
|
|
|
|
|
if record.get("status") == "running":
|
|
|
|
|
synced = await sync_review_style_run_status(record["full_name"])
|
|
|
|
|
out.append(synced)
|
|
|
|
|
else:
|
|
|
|
|
out.append(record)
|
|
|
|
|
return out
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.post("/review-styles")
|
|
|
|
|
async def api_create_review_style(
|
|
|
|
|
body: ReviewStyleCreate,
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
2026-05-21 10:44:04 -07:00
|
|
|
await _require_repo_access_for_user(session["sub"], body.full_name)
|
feat: tune reviewer for precision — web/wiki tools + recalibrated prompt (#1312)
* feat: tune reviewer for precision — web/wiki tools + recalibrated prompt
Reviewer agent now has web_search, fetch_url, and http_request alongside the
finding tools, so it can verify library semantics and consult the DeepWiki
auto-generated wiki for public repos (https://deepwiki.com/<owner>/<repo>)
before flagging cross-file or architectural concerns.
Prompt rewritten to push precision over recall:
- explicit severity ladder pushing reviews toward bimodal high/low instead of
defaulting to medium
- ≤200-char description target (gold set averages ~186 chars; we were at ~436)
- mandatory docs / wiki / code lookup before flagging concurrency, security,
or perf — the three categories that dominated false positives
- "do not flag" list covering compiler/linter-catchable nits, speculative
claims without a concrete attacker/interleaving/scale, style preferences
the codebase doesn't share, and test-quality nits on non-test diffs
- smart file-selection guidance for large PRs (deprioritize generated /
vendored / pure-rename hunks)
Eval config switched to openai:gpt-5.5 + high reasoning effort for the next
benchmark run.
* trim prompt
* subagent prompting
* confidence ratings
* added medium
* enforce confidence threshold
* .
* reviewer: precision-tuned prompt + drop confidence gate
Rewrites the reviewer system prompt around a defensibility bar (anchor +
failure mode + maintainer wouldn't say "not a bug"), an explicit do-not-file
list (style nits, speculation, scope-policing, same-bug fan-out), and a
checklist of 10 bug archetypes drawn from a per-PR audit of the eval golden
set. The audit showed 145 FPs in the last eval split ~28% speculative, ~26%
style-nit, ~31% real-but-unscored (mostly same-archetype fan-out); the new
prompt targets each class directly.
Confidence is still recorded on every finding for post-hoc calibration but
no longer gates publication — the audit showed the gate was a no-op (agent
self-rated 65% of findings "high" regardless), and the prompt's defensibility
bar is the actual discipline. Drops CONFIDENCE_ORDER, CONFIDENCE_THRESHOLD,
the confidence_threshold kwarg on filter_findings_for_publish, the
confidence_filtered score_mode, and the min_confidence kwarg on the eval
target's _extract_comments — all dead once the gate is gone.
Also removes the "informational" severity tier from the Severity enum,
SEVERITY_ORDER, and all validators / tests / docstrings. It was reserved for
FYI observations the dataset never rewards.
* benchmax
* adding google provider
* slight steering
* tuning
* more tuning
* fix
* cleanup
* reducing overfitting
* Add per-repo review style profiles and inject them into the reviewer.
Dashboard users can analyze historical PR review feedback per repository,
edit the resulting style guide, and have it loaded from LangGraph Store at
reviewer runtime (including Martian eval runs) keyed by owner/name.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix review style job errors leaking exception details to clients.
Return generic dashboard messages while logging full stack traces server-side.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-20 11:35:00 -07:00
|
|
|
return await create_review_style(body.full_name, session["sub"])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.get("/review-styles/{full_name:path}")
|
|
|
|
|
async def api_get_review_style(
|
|
|
|
|
full_name: str,
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
|
|
|
|
full_name = normalize_repo_full_name(full_name)
|
|
|
|
|
record = await get_review_style(full_name)
|
|
|
|
|
if not record:
|
|
|
|
|
raise HTTPException(404, "review style not found")
|
|
|
|
|
if record.get("status") == "running":
|
|
|
|
|
record = await sync_review_style_run_status(full_name)
|
|
|
|
|
return record
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.put("/review-styles/{full_name:path}")
|
|
|
|
|
async def api_update_review_style_prompt(
|
|
|
|
|
full_name: str,
|
|
|
|
|
body: ReviewStylePromptUpdate,
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
|
|
|
|
full_name = normalize_repo_full_name(full_name)
|
|
|
|
|
record = await get_review_style(full_name)
|
|
|
|
|
if not record:
|
|
|
|
|
raise HTTPException(404, "review style not found")
|
2026-05-21 10:44:04 -07:00
|
|
|
await _require_repo_access_for_user(session["sub"], full_name)
|
feat: tune reviewer for precision — web/wiki tools + recalibrated prompt (#1312)
* feat: tune reviewer for precision — web/wiki tools + recalibrated prompt
Reviewer agent now has web_search, fetch_url, and http_request alongside the
finding tools, so it can verify library semantics and consult the DeepWiki
auto-generated wiki for public repos (https://deepwiki.com/<owner>/<repo>)
before flagging cross-file or architectural concerns.
Prompt rewritten to push precision over recall:
- explicit severity ladder pushing reviews toward bimodal high/low instead of
defaulting to medium
- ≤200-char description target (gold set averages ~186 chars; we were at ~436)
- mandatory docs / wiki / code lookup before flagging concurrency, security,
or perf — the three categories that dominated false positives
- "do not flag" list covering compiler/linter-catchable nits, speculative
claims without a concrete attacker/interleaving/scale, style preferences
the codebase doesn't share, and test-quality nits on non-test diffs
- smart file-selection guidance for large PRs (deprioritize generated /
vendored / pure-rename hunks)
Eval config switched to openai:gpt-5.5 + high reasoning effort for the next
benchmark run.
* trim prompt
* subagent prompting
* confidence ratings
* added medium
* enforce confidence threshold
* .
* reviewer: precision-tuned prompt + drop confidence gate
Rewrites the reviewer system prompt around a defensibility bar (anchor +
failure mode + maintainer wouldn't say "not a bug"), an explicit do-not-file
list (style nits, speculation, scope-policing, same-bug fan-out), and a
checklist of 10 bug archetypes drawn from a per-PR audit of the eval golden
set. The audit showed 145 FPs in the last eval split ~28% speculative, ~26%
style-nit, ~31% real-but-unscored (mostly same-archetype fan-out); the new
prompt targets each class directly.
Confidence is still recorded on every finding for post-hoc calibration but
no longer gates publication — the audit showed the gate was a no-op (agent
self-rated 65% of findings "high" regardless), and the prompt's defensibility
bar is the actual discipline. Drops CONFIDENCE_ORDER, CONFIDENCE_THRESHOLD,
the confidence_threshold kwarg on filter_findings_for_publish, the
confidence_filtered score_mode, and the min_confidence kwarg on the eval
target's _extract_comments — all dead once the gate is gone.
Also removes the "informational" severity tier from the Severity enum,
SEVERITY_ORDER, and all validators / tests / docstrings. It was reserved for
FYI observations the dataset never rewards.
* benchmax
* adding google provider
* slight steering
* tuning
* more tuning
* fix
* cleanup
* reducing overfitting
* Add per-repo review style profiles and inject them into the reviewer.
Dashboard users can analyze historical PR review feedback per repository,
edit the resulting style guide, and have it loaded from LangGraph Store at
reviewer runtime (including Martian eval runs) keyed by owner/name.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix review style job errors leaking exception details to clients.
Return generic dashboard messages while logging full stack traces server-side.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-20 11:35:00 -07:00
|
|
|
return await set_custom_prompt(full_name, body.custom_prompt)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.post("/review-styles/{full_name:path}/analyze")
|
|
|
|
|
async def api_analyze_review_style(
|
|
|
|
|
full_name: str,
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
|
|
|
|
full_name = normalize_repo_full_name(full_name)
|
2026-05-21 10:44:04 -07:00
|
|
|
token = await _require_repo_access_for_user(session["sub"], full_name)
|
feat: tune reviewer for precision — web/wiki tools + recalibrated prompt (#1312)
* feat: tune reviewer for precision — web/wiki tools + recalibrated prompt
Reviewer agent now has web_search, fetch_url, and http_request alongside the
finding tools, so it can verify library semantics and consult the DeepWiki
auto-generated wiki for public repos (https://deepwiki.com/<owner>/<repo>)
before flagging cross-file or architectural concerns.
Prompt rewritten to push precision over recall:
- explicit severity ladder pushing reviews toward bimodal high/low instead of
defaulting to medium
- ≤200-char description target (gold set averages ~186 chars; we were at ~436)
- mandatory docs / wiki / code lookup before flagging concurrency, security,
or perf — the three categories that dominated false positives
- "do not flag" list covering compiler/linter-catchable nits, speculative
claims without a concrete attacker/interleaving/scale, style preferences
the codebase doesn't share, and test-quality nits on non-test diffs
- smart file-selection guidance for large PRs (deprioritize generated /
vendored / pure-rename hunks)
Eval config switched to openai:gpt-5.5 + high reasoning effort for the next
benchmark run.
* trim prompt
* subagent prompting
* confidence ratings
* added medium
* enforce confidence threshold
* .
* reviewer: precision-tuned prompt + drop confidence gate
Rewrites the reviewer system prompt around a defensibility bar (anchor +
failure mode + maintainer wouldn't say "not a bug"), an explicit do-not-file
list (style nits, speculation, scope-policing, same-bug fan-out), and a
checklist of 10 bug archetypes drawn from a per-PR audit of the eval golden
set. The audit showed 145 FPs in the last eval split ~28% speculative, ~26%
style-nit, ~31% real-but-unscored (mostly same-archetype fan-out); the new
prompt targets each class directly.
Confidence is still recorded on every finding for post-hoc calibration but
no longer gates publication — the audit showed the gate was a no-op (agent
self-rated 65% of findings "high" regardless), and the prompt's defensibility
bar is the actual discipline. Drops CONFIDENCE_ORDER, CONFIDENCE_THRESHOLD,
the confidence_threshold kwarg on filter_findings_for_publish, the
confidence_filtered score_mode, and the min_confidence kwarg on the eval
target's _extract_comments — all dead once the gate is gone.
Also removes the "informational" severity tier from the Severity enum,
SEVERITY_ORDER, and all validators / tests / docstrings. It was reserved for
FYI observations the dataset never rewards.
* benchmax
* adding google provider
* slight steering
* tuning
* more tuning
* fix
* cleanup
* reducing overfitting
* Add per-repo review style profiles and inject them into the reviewer.
Dashboard users can analyze historical PR review feedback per repository,
edit the resulting style guide, and have it loaded from LangGraph Store at
reviewer runtime (including Martian eval runs) keyed by owner/name.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix review style job errors leaking exception details to clients.
Return generic dashboard messages while logging full stack traces server-side.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-20 11:35:00 -07:00
|
|
|
record = await get_review_style(full_name)
|
|
|
|
|
if not record:
|
|
|
|
|
record = await create_review_style(full_name, session["sub"])
|
|
|
|
|
if record.get("status") == "running":
|
2026-05-21 10:44:04 -07:00
|
|
|
record = await sync_review_style_run_status(full_name)
|
|
|
|
|
if record.get("status") == "running":
|
|
|
|
|
raise HTTPException(409, "analysis already running")
|
feat: tune reviewer for precision — web/wiki tools + recalibrated prompt (#1312)
* feat: tune reviewer for precision — web/wiki tools + recalibrated prompt
Reviewer agent now has web_search, fetch_url, and http_request alongside the
finding tools, so it can verify library semantics and consult the DeepWiki
auto-generated wiki for public repos (https://deepwiki.com/<owner>/<repo>)
before flagging cross-file or architectural concerns.
Prompt rewritten to push precision over recall:
- explicit severity ladder pushing reviews toward bimodal high/low instead of
defaulting to medium
- ≤200-char description target (gold set averages ~186 chars; we were at ~436)
- mandatory docs / wiki / code lookup before flagging concurrency, security,
or perf — the three categories that dominated false positives
- "do not flag" list covering compiler/linter-catchable nits, speculative
claims without a concrete attacker/interleaving/scale, style preferences
the codebase doesn't share, and test-quality nits on non-test diffs
- smart file-selection guidance for large PRs (deprioritize generated /
vendored / pure-rename hunks)
Eval config switched to openai:gpt-5.5 + high reasoning effort for the next
benchmark run.
* trim prompt
* subagent prompting
* confidence ratings
* added medium
* enforce confidence threshold
* .
* reviewer: precision-tuned prompt + drop confidence gate
Rewrites the reviewer system prompt around a defensibility bar (anchor +
failure mode + maintainer wouldn't say "not a bug"), an explicit do-not-file
list (style nits, speculation, scope-policing, same-bug fan-out), and a
checklist of 10 bug archetypes drawn from a per-PR audit of the eval golden
set. The audit showed 145 FPs in the last eval split ~28% speculative, ~26%
style-nit, ~31% real-but-unscored (mostly same-archetype fan-out); the new
prompt targets each class directly.
Confidence is still recorded on every finding for post-hoc calibration but
no longer gates publication — the audit showed the gate was a no-op (agent
self-rated 65% of findings "high" regardless), and the prompt's defensibility
bar is the actual discipline. Drops CONFIDENCE_ORDER, CONFIDENCE_THRESHOLD,
the confidence_threshold kwarg on filter_findings_for_publish, the
confidence_filtered score_mode, and the min_confidence kwarg on the eval
target's _extract_comments — all dead once the gate is gone.
Also removes the "informational" severity tier from the Severity enum,
SEVERITY_ORDER, and all validators / tests / docstrings. It was reserved for
FYI observations the dataset never rewards.
* benchmax
* adding google provider
* slight steering
* tuning
* more tuning
* fix
* cleanup
* reducing overfitting
* Add per-repo review style profiles and inject them into the reviewer.
Dashboard users can analyze historical PR review feedback per repository,
edit the resulting style guide, and have it loaded from LangGraph Store at
reviewer runtime (including Martian eval runs) keyed by owner/name.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix review style job errors leaking exception details to clients.
Return generic dashboard messages while logging full stack traces server-side.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-20 11:35:00 -07:00
|
|
|
return await start_review_style_analysis(
|
|
|
|
|
full_name,
|
|
|
|
|
github_token=token,
|
|
|
|
|
created_by=session["sub"],
|
|
|
|
|
)
|
2026-05-21 10:44:04 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.post("/review-styles/{full_name:path}/cancel")
|
|
|
|
|
async def api_cancel_review_style(
|
|
|
|
|
full_name: str,
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> dict[str, Any]:
|
|
|
|
|
del session
|
|
|
|
|
full_name = normalize_repo_full_name(full_name)
|
|
|
|
|
record = await get_review_style(full_name)
|
|
|
|
|
if not record:
|
|
|
|
|
raise HTTPException(404, "review style not found")
|
|
|
|
|
return await cancel_review_style_analysis(full_name)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@router.delete("/review-styles/{full_name:path}")
|
|
|
|
|
async def api_delete_review_style(
|
|
|
|
|
full_name: str,
|
|
|
|
|
session: dict[str, Any] = _SESSION_DEP,
|
|
|
|
|
) -> Response:
|
|
|
|
|
del session
|
|
|
|
|
full_name = normalize_repo_full_name(full_name)
|
|
|
|
|
record = await get_review_style(full_name)
|
|
|
|
|
if not record:
|
|
|
|
|
raise HTTPException(404, "review style not found")
|
|
|
|
|
if record.get("status") == "running":
|
|
|
|
|
await cancel_review_style_analysis(full_name)
|
|
|
|
|
await delete_review_style(full_name)
|
|
|
|
|
return Response(status_code=204)
|