open-swe/INSTALLATION.md
Adam Moussa 1f060f2a1d
Some checks are pending
CI / Lint (push) Waiting to run
CI / Format check (push) Waiting to run
CI / Unit tests (push) Waiting to run
CI / Playwright E2E (push) Waiting to run
chore: sync upstream/main, defer #1621 modular webhooks (#81)
* chore: bake sfw binary into sandbox image (#1611)

sfw only ships a launcher that fetches its real binary at first run and does a
daily update check against api.github.com/repos/SocketDev/sfw-free. Both fail in
the sandbox (restricted egress; the proxy injects the GitHub App installation
token, which lacks access to that repo), so `sfw yarn install` errors with
"could not fetch its binary". Pin sfw 2.0.6, warm + verify the binary cache at
build, and set SFW_SKIP_UPDATE_CHECK=1 so runs use the baked binary offline.

* feat: editable plan mode + fix review-plan banner overlap (#1610)

* feat: editable plan mode + fix review-plan banner overlap

Lets the thread owner edit the plan markdown by hand from the plan-review
page (Edit -> textarea -> Save) via a new PUT /dashboard/api/plan/{id}
endpoint that re-publishes the plan and mirrors it into the sandbox
plan.md, so approve hands the edited plan to the agent as the source of
truth. Also fixes the collapsed git-panel's floating expand button
covering the "Review plan ->" banner by reserving space for it.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: abort plan approval when the published plan read fails

get_plan_content() swallowed store errors and returned None, so a
transient failure during approve would still mark the plan approved and
dispatch the generic fallback text — silently dropping an owner's edited
plan. Read the plan strictly (raise_on_error=True) so approval aborts
instead, matching the comment read.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: show message timestamps (#1609)

* feat: show message timestamps

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: suppress fallback message timestamps

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: stable message + tool-call hover timestamps

Stamp a stable client-side arrival time per message and tool call (keyed
by id, persisted to localStorage). Messages render the timestamp inline;
tool rows reveal a dim timestamp chip on hover. Real backend created_at
still takes precedence when present.

* fix: hide client-stamped message timestamps

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: add PR trace resolution (#1612)

* feat: add PR trace resolution

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: inject reviewer trace context as JSON

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: address review on PR trace resolution

Use the documented LangSmith metadata filter syntax
(and(eq(metadata_key,...), eq(metadata_value,...))) instead of
has(metadata, '{...}'), which does not match runs — _list_thread_runs
was silently returning nothing. Bound full-text searches to a 90-day
window so they don't hit LangSmith's large-window rate limit.

Also folds in the best-effort branch->head-sha resolver (dropping the
weighted scoring/threshold + repo/file evidence + GitHub hydration),
sandbox JSON injection, and the admin "Resolve trace" dry-run endpoint.

The IDOR findings are moot: resolve_pr_to_threads/summarize_agent_session
were removed; resolution now runs deterministically from the trusted run
config with no model-controlled pr_url or thread_id.

* fix: scope branch trace search to the repo

Branch names like fix-tests aren't unique across repos (or older PRs) in
a shared tracing project, so an unscoped branch hit could resolve to an
unrelated thread and write its runs into the reviewer sandbox. Require
the repo slug to co-occur with the branch in matched runs; the full head
SHA stays unscoped since it is globally unique. Addresses open-swe review
on PR #1612.

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: include plan links in PR descriptions (#1613)

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: gate workflow pushes with approval (#1614)

* feat: gate workflow pushes with approval

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: preserve proxy refresh test compatibility

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: bind workflow approvals to pushed ref

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: recover thread work as patch (#1615)

* feat: recover thread work as patch

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: search sandbox cwd for recovery patches

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: omit plan link in PR description when no plan exists (#1618)

Plan links in PR descriptions were always built from the thread id, so
runs that never produced a plan linked to an empty plan-review page.
Now the plan content store is consulted first; the link is only added
when a plan with non-empty markdown actually exists. A transient store
failure degrades gracefully (no link) rather than blocking PR creation.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: add filter & grouping menu to agents threads sidebar (#1617)

Add a Cursor-style control to the agents sidebar that groups (None/Date/
Status/Project), filters (ownership, status, source, pull request, model,
repo, include-resolved), and compacts the threads list. All client-side over
already-fetched sidebar threads; preferences persist in localStorage.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* chore: update langsmith sdk to 0.9.3 (#1616)

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: clickable shared PR header in git panel and reviews (#1620)

* feat: clickable shared PR header in git panel and reviews

Replace the standalone "View PR" button in the agent git panel with a
clickable PR title, matching the reviews view. Extract a shared PrHeader
component reused by both the git panel and the review main body.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* refactor: drop PrHeader wrapper, use shared component directly

The review-side PrHeader was just a thin adapter mapping detail -> the
shared component's props. Inline it at the call site and use the shared
PrHeader directly so there's a single component.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* refactor: durable interrupt dispatch + completion webhook (#1621)

* wip(rebuild): core reliability spine

- remove PR-babysitting (ci_autofix + ci_monitor graph + webhook wiring)
- dispatch core: agent/dispatch.py with multitask_strategy=interrupt +
  durability=sync + completion webhook; reroute all webhook + plan triggers;
  drop the racy in-process lock + is_thread_active busy-check
- completion webhook: agent/completion.py + /webhooks/run-complete loopback
  route for failure/timeout replies (idempotent)

Co-authored-by: open-swe[bot]

* feat(rebuild): async tools, reconcile, shared http timeouts, assembly tuning

Parallel batch on top of the reliability spine:
- async-ify all 24 tools (drop asyncio.run; requests->httpx); re-implement the
  http_request/fetch_url SSRF + DNS-rebinding defense httpx-natively and harden
  the IP check to 'not is_global' (+ IPv4-mapped unwrap)
- reconcile.py: stale pending-run sweep (threads.search -> per-thread runs.list
  -> cancel_many), wired into the scheduler graph via task='reconcile'
- shared DEFAULT_HTTP_TIMEOUT (agent/utils/http.py) on every bare
  httpx.AsyncClient() across utils/dashboard/webapp/middleware
- run budget: MODEL_CALL_RECURSION_LIMIT 5000->250
- fix stale OpenAI->Anthropic fallback id (claude-opus-4-5 -> 4-8)
- drop redundant custom repair middleware (deepagents auto-adds PatchToolCalls)
- confirm tool-result eviction + summarization auto-wired via backend
- slim system prompt ~8% (full harness-profile rewrite deferred)

Co-authored-by: open-swe[bot]

* feat(rebuild): harness-profile prompt + split webhooks out of webapp

- prompt.py: own the system prompt via a registered harness profile
  (OPEN_SWE_SHARED_BASE, kept neutral so the read-only reviewer/analyzer that
  share it stay safe), registered across all 4 providers; per-thread values
  stay in construct_system_prompt. Assembled main-agent prompt ~6.8k -> ~3.1k
  tokens (~55% smaller); de-duped PR/commit/suite/force-push guidance; dropped
  ALL-CAPS markers.
- webapp.py 3325 -> 1890 LOC: moved 14 per-source handlers into
  agent/webhooks/{linear,slack,github}.py; webapp re-exports them for the
  routes + tests; moved handlers reach shared helpers via the webapp namespace
  to preserve the test suite's monkeypatch targets.

Full suite: 1168 passing, lint clean.

Co-authored-by: open-swe[bot]

* Restore MODEL_CALL_RECURSION_LIMIT to 5000 for long-running tasks

Reverts the 250 cap from the run-budget change — long-running tasks legitimately
need many model calls. The notify_step_limit_reached safety net still fires if a
run does hit the cap, so runs end with a signal either way.

Co-authored-by: open-swe[bot]

* fix: address PR review (auth, SSRF, interrupted status, redirect headers)

- completion.py: drop `interrupted` from failure statuses — with
  multitask_strategy=interrupt a follow-up ends the prior run as interrupted,
  which is healthy, not a failure to report. [open-swe]
- /webhooks/run-complete: shared-secret auth — dispatch appends ?token= when
  RUN_COMPLETE_WEBHOOK_SECRET is set; route verifies via hmac.compare_digest.
  [corridor-security]
- SSRF: extract the URL validator to agent/utils/url_safety.py and apply it
  before server-side image fetches in multimodal.fetch_image_block.
  [corridor-security]
- http_request: preserve caller headers/extensions across redirect hops instead
  of dropping them on the first hop. [open-swe]

Co-authored-by: open-swe[bot]

* chore: remove REBUILD_PLAN.md (planning doc, not needed in the repo)

Co-authored-by: open-swe[bot]

* fix: fail closed on run-complete webhook auth when secret unset

Corridor follow-up: verify_run_complete_token returns False (not True) when
RUN_COMPLETE_WEBHOOK_SECRET is unset, so the public route is never
unauthenticated. Logs a startup warning when the secret is absent, and dispatch
skips registering the webhook when there's no secret (no rejected callbacks).

Co-authored-by: open-swe[bot]

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: restore forced tool call to prevent premature run stops (#1622)

Restore the ensure_no_empty_msg middleware and the always-call-a-tool system-prompt instruction that #1535 removed. When the model emits a message with no tool call (and hasn't already messaged the user or confirmed completion), the middleware re-injects a no_op / confirming_completion tool call so the run continues instead of ending mid-task.

Shipping to test whether it fixes runs that stop halfway through.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* chore(deps): bump langgraph-checkpoint from 4.1.0 to 4.1.1 (#1619)

Bumps [langgraph-checkpoint](https://github.com/langchain-ai/langgraph) from 4.1.0 to 4.1.1.
- [Release notes](https://github.com/langchain-ai/langgraph/releases)
- [Commits](https://github.com/langchain-ai/langgraph/compare/checkpoint==4.1.0...checkpoint==4.1.1)

---
updated-dependencies:
- dependency-name: langgraph-checkpoint
  dependency-version: 4.1.1
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* fix: post reviewer resolution notes verbatim (#1624)

* fix: post reviewer resolution notes verbatim

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: stabilize dashboard follow-up e2e

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: preserve dashboard attribution in e2e

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: make e2e attribution marker durable

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: only echo found e2e attribution

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: check live dashboard attribution in e2e

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* hotfix: stop prompting agent/reviewer to wrap installs in sfw (#1625)

Installs hung when prefixed with sfw inside the sandbox (trace 019f0608
stalled on a pending `sfw npm install` execute, never returned). Strip the
Socket Firewall guidance from the agent and reviewer prompts so installs run
through the project's package manager directly. sfw stays in the Docker image;
nothing invokes it now.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: make plan view mobile friendly (#1636)

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: fall back to vision model for image threads (#1626)

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: surface Slack thread errors (#1627)

* fix: surface Slack thread errors

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: don't set failure_reply_posted on Slack preprocessing errors

The preprocessing error handler was setting failure_reply_posted=True,
the same idempotency flag handle_run_completion checks to suppress
duplicate run-failure replies. Since preprocessing failures happen
before any run exists but the flag persists on the thread, a subsequent
run failure on the same thread would be silently ignored.

The preprocessing handler already posts its own Slack reply, so the
run-completion idempotency flag should not be set here.

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* chore: avoid recapping Slack replies (#1629)

* chore: avoid recapping Slack replies

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* chore: simplify Slack reply prompt wording

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

---------

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: update Slack trace reply on web handoff (#1630)

* fix: update Slack trace reply on web handoff

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: trigger web handoff on dashboard starts

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: format web handoff as contextual fragment

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: preserve trace_message_ts when overwriting Slack run mapping

When store_slack_run_mapping is called without trace_message_ts (e.g. on
follow-up Slack mentions), it was unconditionally overwriting the
thread-level mapping and clobbering the timestamp captured from the
initial trace reply. After that, _notify_slack_web_handoff could not find
the original message, so a subsequent move to Web silently skipped the
Slack trace update.

Now, when trace_message_ts is not passed, the existing thread mapping is
read first and its trace_message_ts is preserved.

* style: ruff format

---------

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Johannes du Plessis <51395795+johannes117@users.noreply.github.com>
Co-authored-by: Johannes du Plessis <johannes@langchain.dev>

* fix: pre-bundle shiki/@pierre deps to stop dev dynamic-import failures (#1643)

* fix(ui): pre-bundle shiki/@pierre deps to stop dev dynamic-import failures

shiki lazy-imports a grammar per language and these libs only live inside
lazy route components, so Vite's startup scanner never sees them. They get
discovered on first thread navigation, triggering a dep re-optimize +
force-reload that aborts the in-flight route-chunk import, surfacing as
"Failed to fetch dynamically imported module: .../$threadId.tsx".

Pre-bundle them (and the github themes + common code-block languages) via
optimizeDeps.include so the optimize happens once at startup. Dev-only;
production bundles are unaffected.

* fix: pre-bundle canonical shiki docker/make langs instead of aliases

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: show queued dashboard follow-ups (#1631)

* feat: show queued dashboard follow-ups

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: de-dupe queued follow-ups while streaming

---------

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Johannes du Plessis <johannes@langchain.dev>

* feat: notify Slack on plan approval (#1632)

* feat: notify Slack on plan approval

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: post Slack approval notice after successful dispatch

Move the _maybe_post_plan_approved_to_slack call until after
_dispatch_followup succeeds so the Slack thread is not told
implementation is beginning before the LangGraph run is created.

Addresses PR review comment.

---------

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Johannes du Plessis <johannes@langchain.dev>

* feat: include Slack channel context in prompts (#1633)

Add cached Slack channel metadata enrichment for Slack-triggered runs so prompts can include channel names and descriptions without duplicate conversations.info calls.\n\nCo-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>

* chore: keep plan guidance high-level (#1634)

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: publish plans from sandbox files (#1635)

* feat: publish plans from sandbox files

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: avoid fixed plan filenames

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: virtualize local sandbox file paths

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: preserve plan_file_path across set_plan_status

set_plan_status was rewriting the content record with only markdown
and status, dropping plan_file_path. After a reject, the owner's
dashboard edit would mirror to a different file than the agent's
original, and the next save_plan could republish the stale file.
Preserve plan_file_path when updating status.

---------

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: return to thread after plan approval (#1637)

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: add Slack breakout thread tool (#1638)

* feat: add Slack breakout thread tool

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* chore: make fake LLM scripts declarative

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: exclude slack_start_new_thread from plan mode

The breakout tool can dispatch a fresh agent run that starts outside the
current plan-mode state, bypassing the approval flow. Add it to
PLAN_MODE_EXCLUDED_TOOLS so it's hidden alongside the other mutating
tools while planning.

---------

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* chore: require bun for ui agent work (#1639)

Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: request actions read for sandbox logs (#1642)

* fix: request actions read for sandbox logs

Request optional Actions read permission for sandbox proxy tokens, with fallback for installations that have not approved it yet. Update setup docs and prompt guidance for safe GitHub Actions log usage.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: restore actions:read scope after workflow push

After an approved workflow push, the guard was restoring the proxy with
BASE_RUNTIME_PROXY_TOKEN_PERMISSIONS, which excludes the actions: read
scope this PR adds. Restore with RUNTIME_PROXY_TOKEN_PERMISSIONS (which
includes actions: read) and fall back to BASE if the install hasn't
granted Actions read — mirroring the pattern in _create_sandbox_with_proxy.

Addresses review comment on PR #1642.

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* fix: widen split review diffs (#1647)

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* chore: install missing deps before verification (#1646)

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* chore: switch ui to pnpm (#1645)

* chore: require pnpm for ui agent work

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* chore: switch ui to pnpm

Replace Bun and Yarn lockfiles with pnpm lockfile and update UI/Vercel commands to use pnpm.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* ci: use corepack for ui pnpm e2e build

Run pnpm through Corepack in the E2E global setup so CI can use the pinned package manager without a separate pnpm install step.

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* feat: add Sonnet 5 to model picker (#1651)

* chore: update Sonnet examples to Sonnet 5

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* chore: add Sonnet 5 to model picker

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>

* Remove dead breakout-thread e2e scenario after dropping the tool

The merge resolution deferred upstream's Slack breakout-thread tool
(slack_start_new_thread, #1638) since it depends on the #1621 dispatch
module, but the e2e harness still scripted it. Removing the tool name
from fake_llm.py's _tool_step call left a malformed scenario, crashing
the langgraph-dev web server at import (TypeError: _tool_step() missing
'call_id') and failing Playwright E2E.

Drop the "breakout" script scenario, its _is_breakout_request helper +
ScriptRule, and the corresponding full_flow.spec.ts test.

* Revert upstream pnpm switch; keep bun for the UI build

The merge auto-adopted upstream's pnpm switch (#1645) in tests/e2e/
global-setup.ts and ui/package.json, but our fork builds the UI with
bun (vercel.json + the E2E workflow's setup-bun). That left the
Playwright globalSetup running `corepack pnpm install --frozen-lockfile`
with no pnpm-lock.yaml, failing E2E at UI build time.

Revert global-setup.ts and ui/package.json to the dev (bun) baseline,
drop the merge-added ui/pnpm-lock.yaml, and remove the re-added
ui/AGENTS.md (our fork had deleted it).

* Align plan-review e2e + UI with the HEAD (pre-#1635) backend

The merge left a split plan vertical: the backend save_plan/plan_api are
HEAD (we deferred the editable-plan/sandbox-publish features #1610/#1635/
#1637 per #80), but the plan UI and e2e harness were upstream's. The
fake_llm scenario called save_plan(plan_file_path=...) — upstream's
file-based #1635 contract — while HEAD save_plan takes plan_markdown,
so the plan never saved and PlanReview never rendered (E2E failure on
the plan-review locator).

Pass plan_markdown to save_plan, and revert PlanReview.tsx / plan.ts /
$threadId_.plan.tsx / plan_review.spec.ts to the dev baseline so the
whole plan flow (save -> render -> approve -> implement) is consistent
with the HEAD backend.

---------

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: Johannes du Plessis <johannes@langchain.dev>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Co-authored-by: Ramon Nogueira <ramon.nogueira@langchain.dev>
Co-authored-by: Ramon Nogueira <270434257+ramon-langchain@users.noreply.github.com>
Co-authored-by: Caroline di Vittorio <43390382+carolinedivittorio@users.noreply.github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Johannes du Plessis <51395795+johannes117@users.noreply.github.com>
Co-authored-by: Ankush Gola <9536492+agola11@users.noreply.github.com>
Co-authored-by: Mukil Loganathan <mukil@langchain.dev>
2026-06-30 16:45:19 -04:00

43 KiB
Raw Blame History

Installation Guide

This guide walks you through setting up Open SWE end-to-end: local development, GitHub App creation, LangSmith configuration, webhooks, the web dashboard, and production deployment.

Open SWE has two runnable pieces:

  • The backend — a LangGraph app (three graphs: agent, reviewer, analyzer) plus a FastAPI app (agent.webapp:app) that owns the webhooks and the dashboard API. Both are served together by langgraph dev.
  • The dashboard — a TanStack Start + Vite web app in ui/ (package name open-swe-dashboard). It's a thin client over the FastAPI dashboard API (/dashboard/api/*): GitHub-login, per-user model/profile settings, team defaults, enabled-repo and review-style management, user mappings, and the Agents chat UI. It's optional for pure webhook-driven use, but recommended.

The steps are ordered to avoid forward references. Each step only depends on things you've already completed.

Prerequisites

  • Python 3.11 – 3.13 (3.14 is not yet supported due to dependency constraints)
  • uv package manager
  • LangGraph CLI
  • ngrok (for local development — exposes webhook endpoints to the internet)
  • pnpm (only if you want to run the dashboard UI locally — see step 8). Node 20+ also works, but ui/pnpm-lock.yaml is the canonical lockfile.

1. Clone and install

git clone https://github.com/langchain-ai/open-swe.git
cd open-swe
uv venv
source .venv/bin/activate
uv sync --all-extras

2. Start ngrok

You'll need the ngrok URL in subsequent steps when configuring webhooks, so start it first.

ngrok http 2024 --url https://some-url-you-configure.ngrok.dev

You don't need to pass the --url flag, however doing so will use the same subdomain each time you startup the server. Without this, you'll need to update the webhook URL in GitHub, Slack and Linear every time you restart your server for local development.

Copy the HTTPS URL you set, or if you didn't pass --url, the one ngrok gives you. You'll paste this into the webhook settings in steps 3 and 5.

Keep this terminal open — ngrok needs to stay running during local development. Use a second terminal for the rest of the steps.

3. Create a GitHub App

Open SWE authenticates as a GitHub App to clone repos, push branches, and open PRs.

3a. Choose your OAuth provider ID

Before creating the app you need to decide on an OAuth provider ID — this is a short string you'll use in both GitHub and LangSmith to link the two. Pick something memorable, for example:

your-org-github-oauth

Write this down. You'll use it in the callback URL below and again in step 4 when configuring LangSmith.

3b. Create the app

  1. Go to GitHub Settings → Developer settings → GitHub Apps → New GitHub App
  2. Fill in:
    • App name: open-swe (or your preferred name)
    • Homepage URL: This can be any valid URL — it's only shown on the GitHub Marketplace page (which you won't be using). Use something like https://github.com/langchain-ai/open-swe
    • Callback URL: GitHub Apps allow multiple callback URLs (one per line). Add both:
      1. https://smith.langchain.com/host-oauth-callback/<your-provider-id> — replace <your-provider-id> with the ID you chose in step 3a (e.g. https://smith.langchain.com/host-oauth-callback/your-org-github-oauth). This is the agent-runtime OAuth callback, brokered by LangSmith (step 4b).
      2. http://localhost:2024/dashboard/api/auth/callback — the dashboard-login OAuth callback (step 8). For production, also add https://<your-dashboard-api-url>/dashboard/api/auth/callback. This is a separate, direct GitHub OAuth flow (not via LangSmith), so it needs its own callback URL.
    • Request user authorization (OAuth) during installation: ✅ Enable this
    • Webhook URL: https://<your-ngrok-url>/webhooks/github — use the ngrok URL from step 2
    • Webhook secret: generate one and save it — you'll need it later as GITHUB_WEBHOOK_SECRET:
      openssl rand -hex 32
      
  3. Set permissions:
    • Repository permissions:
      • Contents: Read & write
      • Pull requests: Read & write
      • Issues: Read & write
      • Checks: Read & write — reports an "Open SWE Review" check run on PRs while an auto-review runs, and reads third-party CI conclusions for the auto-fix flow (it watches failing checks on agent-authored PRs and pushes fixes). Without it, check-run creation fails (logged, best-effort) but reviews still work, and CI auto-fix is disabled.
      • Commit statuses: Read-only — only needed if you enable the Status event below; the CI auto-fix flow reads the legacy combined commit-status API for integrations that report via statuses instead of check runs. Without it, status-based CI is silently ignored (logged as "Failed to read combined status").
      • Actions: Read-only — optional; lets Open SWE's sandbox proxy tokens download GitHub Actions workflow/job logs when troubleshooting CI failures. Do not grant Actions write for log access: write permission also allows rerunning, canceling, and deleting workflow runs, which is unnecessary for diagnostics.
      • Workflows: Read & write — required to let Open SWE push branches containing GitHub Actions workflow changes after explicit human approval. Runtime sandbox tokens are still minted without this permission by default and are elevated only around an approved workflow push.
      • Metadata: Read-only
    • Organization permissions (required only if you plan to set ALLOWED_GITHUB_ORGS — see step 5 / Security):
      • Members: Read-only — used to verify org membership for the dashboard-login gate via GET /orgs/{org}/memberships/{username}. Without this permission that call returns 403, the check fails closed, and every dashboard login is rejected.
  4. Under Subscribe to events, enable:
    • Issue comment
    • Pull request review
    • Pull request review comment
    • Check run — required for CI auto-fix (watching failing GitHub Actions checks on agent PRs)
    • Check suite — required for CI auto-fix
    • Workflow run — required for CI auto-fix
    • Status — optional; covers integrations that report via the legacy commit-status API
  5. Click Create GitHub App

3c. Collect credentials

After creating the app:

  1. App ID — shown at the top of the app's settings page. Save this as GITHUB_APP_ID.
  2. Private key — scroll down to Private keys → click Generate a private key. A .pem file will download. Save its contents as GITHUB_APP_PRIVATE_KEY.
  3. Client ID — shown near the top of the app's settings page (starts with Iv...). Save this as GITHUB_APP_CLIENT_ID.
  4. Client secret — under Client secrets → Generate a new client secret. Save it as GITHUB_APP_CLIENT_SECRET.

GITHUB_APP_CLIENT_ID / GITHUB_APP_CLIENT_SECRET power the dashboard login flow (the direct GitHub OAuth in 3b's second callback URL). They are independent of the LangSmith OAuth provider in step 4b — the dashboard talks to GitHub directly, while the agent runtime resolves per-user tokens through LangSmith.

3d. Install the app on your repositories

  1. From your app's settings page, click Install App in the sidebar
  2. Select your org or personal account
  3. Choose which repositories Open SWE should have access to
  4. Click Install
  5. After installation, look at the URL in your browser — it will look like:
    https://github.com/settings/installations/12345678
    
    or for an org:
    https://github.com/organizations/YOUR-ORG/settings/installations/12345678
    
    The number at the end (12345678) is your Installation ID. Save this as GITHUB_APP_INSTALLATION_ID.

Note

: The installation page may prompt you to authenticate with LangSmith. If you haven't set up LangSmith yet (step 4), that's fine — you can still grab the Installation ID from the URL and complete the OAuth setup later.

4. Set up LangSmith

Open SWE uses LangSmith for:

  • Tracing: all agent runs are logged for debugging and observability
  • Sandboxes: each task runs in an isolated LangSmith cloud sandbox

4a. Get your API key, project and tenant IDs

  1. Create a LangSmith account if you don't have one
  2. Go to Settings → API Keys → Create API Key
  3. Save it as LANGSMITH_API_KEY_PROD
  4. Get your Tenant ID: Visit LangSmith, login, then copy the UUID in the URL. Example: if your URL is https://smith.langchain.com/o/72184268-01ea-4d29-98cc-6cfcf0f2abb0/agents/chat -> the tenant ID would be 72184268-01ea-4d29-98cc-6cfcf0f2abb0. Save it as LANGSMITH_TENANT_ID_PROD.
  5. Get your Project ID: open your tracing project in LangSmith, then click on the ID button in the top left, directly next to the project name. Save it as LANGSMITH_TRACING_PROJECT_ID_PROD

Note on per-graph tracing projects. The graphs trace into separate projects by name — open-swe-agent (main agent) and open-swe-review (reviewer/analyzer). "View trace" links resolve the correct project ID from these names automatically (via the LANGSMITH_API_KEY_PROD client), so make sure projects with these names exist in your tenant. If a name can't be resolved, links fall back to LANGSMITH_TRACING_PROJECT_ID_PROD, so set it to whichever project you want links to point at by default.

This is the agent-runtime OAuth provider: it lets each agent run authenticate with the triggering user's own GitHub account, brokered by LangSmith. (It is separate from the dashboard-login OAuth, which uses GITHUB_APP_CLIENT_ID/GITHUB_APP_CLIENT_SECRET directly — see step 3c.) Without it, all agent operations use the GitHub App's installation token (a shared bot identity).

What this affects:

  • With per-user OAuth: PRs and commits show the triggering user's identity; each user's GitHub permissions are respected
  • Without it (bot-token-only mode): all PRs and commits appear as the GitHub App bot; the app's installation-level permissions are used for everything

To set up per-user OAuth:

  1. In LangSmith, go to Settings → OAuth Providers → Add Provider
  2. Set the Provider ID to the same string you chose in step 3a (e.g. your-org-github-oauth)
  3. Enter the Client ID and Client Secret from your GitHub App (found on the GitHub App settings page under OAuth credentials)
  4. Enter the Authorization URL as https://github.com/login/oauth/authorize and the Token URL as https://github.com/login/oauth/access_token.
  5. Leave "Enable PKCE" unchecked.
  6. Save. You'll reference this Provider ID as GITHUB_OAUTH_PROVIDER_ID in your environment variables.

4c. Sandbox snapshots

LangSmith sandboxes provide the isolated execution environment for each agent run. Open SWE boots each sandbox from a pre-built snapshot — you build the snapshot once (from a Docker image) and then reference it by UUID.

(Optional) Build and Push a custom Docker Image to Docker hub First build and push the sandbox Docker image to a registry LangSmith can pull from. On Apple Silicon, force linux/amd64

docker buildx build \
  --platform linux/amd64 \
  -t <your-docker-hub>/<name-of-your-image> \
  --push .

For a multi-arch tag that also runs locally on Apple Silicon:

docker buildx build \
  --platform linux/amd64,linux/arm64 \
  -t <your-docker-hub>/<name-of-your-image> \
  --push .

Then build a snapshot in the LangSmith UI (Sandboxes → Snapshots → New), or via the SDK:

from langsmith.sandbox import SandboxClient

client = SandboxClient(api_key="<your key>")
snapshot = client.create_snapshot(
    name="open-swe",
    docker_image="johanneslangchain/open-swe-sandbox:gh-cli-amd64",  # built from ./Dockerfile
    fs_capacity_bytes=32 * 1024**3,
)
print(snapshot.id)

You can also use the helper script:

uv run python scripts/create_sandbox_snapshot.py \
  --name open-swe-gh-cli-amd64 \
  --image johanneslangchain/open-swe-sandbox:gh-cli-amd64

Then set the resulting UUID in your environment:

DEFAULT_SANDBOX_SNAPSHOT_ID="<snapshot-uuid>"
# Optional; overrides the snapshot's root FS size at sandbox boot. Default is 32 GiB.
DEFAULT_SANDBOX_SNAPSHOT_FS_CAPACITY_BYTES="34359738368"
# Optional; number of vCPUs per sandbox. Default is 4.
DEFAULT_SANDBOX_VCPUS="4"
# Optional; memory in bytes per sandbox. Default is 15 GiB.
DEFAULT_SANDBOX_MEM_BYTES="16106127360"
# Optional; auto-stop a sandbox after this many seconds of inactivity. Default is 7200 (2 hours). 0 disables.
DEFAULT_SANDBOX_IDLE_TTL_SECONDS="7200"
# Optional; delete a stopped sandbox after this many seconds. Default is 86400 (24 hours). 0 disables.
DEFAULT_SANDBOX_DELETE_AFTER_STOP_SECONDS="86400"
# Optional; required only for the admin Repository Snapshots page/template generator.
REPO_SNAPSHOT_BASE_IMAGE="<your-docker-hub>/<name-of-your-image>"

DEFAULT_SANDBOX_SNAPSHOT_ID is required when SANDBOX_TYPE=langsmith. The server validates this at startup and refuses to boot if it's missing. The snapshot should include the GitHub CLI from the project Dockerfile; Open SWE authenticates git and gh through the LangSmith sandbox proxy using runtime-minted GitHub App installation tokens, not deployment-stored GitHub access tokens.

REPO_SNAPSHOT_BASE_IMAGE should point at the same published Open SWE sandbox image you used to create the default snapshot (for example, the image built from ./Dockerfile). The admin Repository Snapshots page uses it as the FROM line when generating per-repo Dockerfile templates. If it is not set, template generation is intentionally disabled so admins do not accidentally build repo-scoped snapshots from a bare image that lacks Open SWE's required tools (git, gh, sfw, language runtimes, and proxy assumptions).

5. Set up triggers

Open SWE can be triggered from GitHub, Linear, and/or Slack. Configure whichever surfaces your team uses — you don't need all of them.

GitHub

GitHub triggering works automatically once your GitHub App is set up (step 3). Users can:

  • Tag @openswe in issue titles or bodies to start a task
  • Tag @openswe in issue comments for follow-up instructions
  • Tag @openswe in PR review comments to have it address review feedback

Which GitHub users can trigger the agent is controlled by the user mapping (GitHub login ⇄ work email ⇄ optional Slack ID), stored in the LangGraph Store rather than in code. Manage it in the dashboard under Admin → User mappings:

  • Add / update a single mapping (GitHub login + work email, plus an optional Slack user ID). The list is paged (20 per page).
  • Users can also self-onboard: when an unmapped person tags Open SWE in Slack, the agent runs with limited (GitHub App installation) permissions and posts a "link your GitHub account" prompt. Completing the org-gated GitHub OAuth login records a self mapping (carrying the originating Slack ID and work email). Self-signup is therefore bounded by the same ALLOWED_GITHUB_ORGS gate as dashboard login.

You should also configure which GitHub organizations and/or repositories the agent is allowed to operate on. You can specify allowed orgs, specific owner/repo pairs, or both:

# Allow all repos in these orgs
ALLOWED_GITHUB_ORGS="langchain-ai,anthropics"

# Allow specific repos (owner/repo format)
ALLOWED_GITHUB_REPOS="some-user/their-repo,another-org/specific-repo"

A GitHub or Linear webhook is accepted if the resolved repo's org is in ALLOWED_GITHUB_ORGS or the owner/repo is in ALLOWED_GITHUB_REPOS. If both are empty, all repos are allowed. Slack mentions are not rejected from regex-inferred repository text; repository access is bounded by the GitHub App installation permissions.

ALLOWED_GITHUB_ORGS also gates dashboard login: when set, only GitHub accounts that are active members of one of the listed organizations can complete the OAuth login and receive a session. Membership is verified server-side with the GitHub App installation token (so private memberships are visible and no extra OAuth scope is required), and the check fails closed on any API error. When ALLOWED_GITHUB_ORGS is empty, dashboard login is open to any GitHub account (the prior behavior).

Required GitHub App permission: the membership check calls GET /orgs/{org}/memberships/{username}, which requires the GitHub App's Organization → Members: Read-only permission (see step 3b). If you set ALLOWED_GITHUB_ORGS without granting that permission, the call returns 403, the check fails closed, and every dashboard login is rejected. After changing an installed app's permissions, GitHub requires you to approve the new permission on each installation before it takes effect.

Linear (optional)

Open SWE listens for Linear comments that mention @openswe.

Create a webhook:

  1. In Linear, go to Settings → API → Webhooks → New webhook
  2. Fill in:
    • Label: open-swe
    • URL: https://<your-ngrok-url>/webhooks/linear — use the ngrok URL from step 2
    • Secret: generate with openssl rand -hex 32 — save this as LINEAR_WEBHOOK_SECRET
  3. Under Data change events, enable Comments → Create only
  4. Click Create webhook

Get your API key:

  1. Go to Settings → API → Personal API keys → New API key
  2. Name it open-swe, select All access, and copy the key
  3. Save it as LINEAR_API_KEY

Configure team-to-repo mapping:

Open SWE routes Linear issues to GitHub repos based on the Linear team and project. Edit the mapping in agent/utils/linear_team_repo_map.py:

LINEAR_TEAM_TO_REPO = {
    "My Team": {"owner": "my-org", "name": "my-repo"},
    "Engineering": {
        "projects": {
            "backend": {"owner": "my-org", "name": "backend"},
            "frontend": {"owner": "my-org", "name": "frontend"},
        },
        "default": {"owner": "my-org", "name": "monorepo"},
    },
}

Users can also override the team/project mapping per-comment by including repo:owner/name (or a GitHub URL) in their @openswe comment. The mapping is used as a fallback when no repo is specified in the comment text.

Slack (optional)

Create a Slack App:

  1. Go to api.slack.com/apps → Create New App → From a manifest
  2. Copy the manifest below, replacing the two placeholder URLs:
    • Replace <your-provider-id> with the OAuth provider ID from step 3a
    • Replace <your-ngrok-url> with the backend URL from step 2 (or your deployed LangGraph/FastAPI URL in production)
Slack App Manifest
{
    "display_information": {
        "name": "Open SWE",
        "description": "Enables Open SWE to interact with your workspace",
        "background_color": "#000000"
    },
    "features": {
        "app_home": {
            "home_tab_enabled": false,
            "messages_tab_enabled": true,
            "messages_tab_read_only_enabled": false
        },
        "bot_user": {
            "display_name": "Open SWE",
            "always_online": true
        }
    },
    "oauth_config": {
        "redirect_urls": [
            "https://smith.langchain.com/host-oauth-callback/<your-provider-id>",
            "http://localhost:2024/dashboard/api/slack/callback"
        ],
        "scopes": {
            "bot": [
                "reactions:write",
                "app_mentions:read",
                "channels:history",
                "channels:read",
                "chat:write",
                "groups:history",
                "groups:read",
                "im:history",
                "im:read",
                "im:write",
                "mpim:history",
                "mpim:read",
                "team:read",
                "users:read",
                "users:read.email"
            ]
        }
    },
    "settings": {
        "event_subscriptions": {
            "request_url": "https://<your-ngrok-url>/webhooks/slack",
            "bot_events": [
                "app_mention",
                "message.im",
                "message.mpim"
            ]
        },
        "interactivity": {
            "is_enabled": true,
            "request_url": "https://<your-ngrok-url>/webhooks/slack/interactivity"
        },
        "org_deploy_enabled": false,
        "socket_mode_enabled": false,
        "token_rotation_enabled": false
    }
}
  1. Install the app to your workspace and copy the Bot User OAuth Token (xoxb-...)

Slack URL checklist:

Both Slack URLs must point at the Open SWE backend that serves agent.webapp:app (locally, your ngrok URL forwarding to langgraph dev; in production, your LangGraph/FastAPI deployment URL), not the dashboard frontend URL.

  • Event Subscriptions → Request URL: https://<your-backend-url>/webhooks/slack
  • Interactivity & Shortcuts → Interactivity Request URL: https://<your-backend-url>/webhooks/slack/interactivity

Slack Block Kit option buttons only work when Interactivity is enabled and pointed at /webhooks/slack/interactivity.

Credentials you'll need:

  • SLACK_BOT_TOKEN: the Bot User OAuth Token (xoxb-...)
  • SLACK_SIGNING_SECRET: found under Basic Information → App Credentials
  • SLACK_BOT_USER_ID: the bot's user ID (find it in Slack by clicking the bot's profile)
  • SLACK_BOT_USERNAME: the bot's display name (e.g. open-swe)

Default repo:

Slack messages are routed to the Slack default repo (SLACK_REPO_OWNER/SLACK_REPO_NAME, falling back to DEFAULT_REPO_OWNER/DEFAULT_REPO_NAME — see step 6) unless the user specifies one with repo:owner/name in their message.

"Sign in with Slack" account linking (optional):

The dashboard can let a user link their Slack identity to their GitHub login via Slack OIDC ("Sign in with Slack"). This is what lets a Slack-triggered run resolve to the right GitHub user. To enable it:

  1. The manifest above already registers the OIDC redirect (.../dashboard/api/slack/callback). Under OpenID Connect (or Sign in with Slack) make sure the openid, email, and profile user scopes are available.
  2. From Basic Information → App Credentials, save the app's Client ID as SLACK_CLIENT_ID and Client Secret as SLACK_CLIENT_SECRET.
  3. (Optional) Set SLACK_TEAM_ID (your workspace ID, T...) to restrict linking to a single workspace.

If SLACK_CLIENT_ID/SLACK_CLIENT_SECRET are unset, the "Sign in with Slack" link is simply disabled; the rest of Slack triggering still works.

6. Environment variables

Create a .env file in the project root. Below is the full list — only fill in the sections relevant to the triggers you configured.

# === LangSmith ===
LANGSMITH_API_KEY_PROD=""              # From step 4a
LANGCHAIN_TRACING_V2="true"
LANGCHAIN_PROJECT=""                   # LangSmith project name for traces
LANGSMITH_TENANT_ID_PROD=""           
LANGSMITH_TRACING_PROJECT_ID_PROD=""   # Fallback project ID for "View trace" links; graphs trace into the open-swe-agent / open-swe-review projects by name
LANGSMITH_URL_PROD="https://smith.langchain.com"                 

# === LLM ===
ANTHROPIC_API_KEY=""                   # Anthropic API key
OPENAI_API_KEY=""                      # OpenAI API key (when using openai: models)
GOOGLE_API_KEY=""                      # Google AI API key (when using google_genai: models)
FIREWORKS_API_KEY=""                   # Fireworks API key (when using fireworks: models)

# === GitHub App (required) ===
GITHUB_APP_ID=""                       # From step 3c
GITHUB_APP_PRIVATE_KEY="-----BEGIN RSA PRIVATE KEY-----
...
-----END RSA PRIVATE KEY-----
"
GITHUB_APP_INSTALLATION_ID=""          # From step 3d

# === GitHub Webhook (required) ===
GITHUB_WEBHOOK_SECRET=""               # The secret you generated in step 3b

# === Dashboard GitHub OAuth (required for the dashboard) ===
# Direct GitHub OAuth used by the dashboard login flow (not via LangSmith).
GITHUB_APP_CLIENT_ID=""                # From step 3c
GITHUB_APP_CLIENT_SECRET=""            # From step 3c

# === Agent-runtime GitHub OAuth via LangSmith (optional) ===
# Without these, all agent operations use the GitHub App's bot token.
# With these, each agent run authenticates as the triggering user.
GITHUB_OAUTH_PROVIDER_ID=""            # The provider ID from steps 3a / 4b
# Secret used to mint short-lived service JWTs that ask LangSmith to resolve a
# specific user's GitHub token. Needed for per-user token resolution in deployed mode.
X_SERVICE_AUTH_JWT_SECRET=""

# === Repo Allowlist (optional) ===
# Comma-separated list of GitHub orgs the agent is allowed to operate on.
# Also gates dashboard login to members of these orgs (requires the GitHub App's
# Organization -> Members: Read-only permission; without it, all dashboard logins are rejected).
# Leave empty to allow all orgs.
ALLOWED_GITHUB_ORGS=""                 # e.g. "my-org,my-other-org"
# Comma-separated list of specific owner/repo pairs the agent is allowed to operate on.
# For GitHub/Linear webhooks, a repo is allowed if its org is in ALLOWED_GITHUB_ORGS OR its owner/repo is in ALLOWED_GITHUB_REPOS.
# Slack mentions are not rejected from regex-inferred repository text; repository access is bounded by GitHub App installation permissions.
# Leave both empty to allow all repos.
ALLOWED_GITHUB_REPOS=""                # e.g. "some-user/their-repo,another-org/specific-repo"

# === Default Repository ===
# Used across all triggers when no repo is specified.
DEFAULT_REPO_OWNER=""                  # Default GitHub org (e.g. "my-org")
DEFAULT_REPO_NAME=""                   # Default GitHub repo (e.g. "my-repo")

# === Dashboard (required to run the web dashboard) ===
# Public URL that browsers use for /dashboard/api/* and OAuth callbacks.
# Use the FastAPI backend URL for local/cross-origin direct API calls.
# Use the dashboard frontend URL when a same-origin frontend rewrite proxies /dashboard/api/*.
# Its scheme drives cookie security: http:// => SameSite=Lax (local);
# https:// => Secure + SameSite=None (production).
DASHBOARD_API_BASE_URL="http://localhost:2024"
# Public base URL of the dashboard frontend (the ui/ app). Default post-login redirect.
DASHBOARD_BASE_URL="http://localhost:3000"
# HMAC secret for dashboard JWTs (session cookie and OAuth state).
DASHBOARD_JWT_SECRET=""                # Generate with: openssl rand -hex 32
# Comma-separated origins allowed for credentialed CORS and post-login redirects.
# Required whenever the frontend and API are on different origins — including local
# dev (UI :3000 -> API :2024 is cross-origin). CORS is only enabled when this is set.
DASHBOARD_ALLOWED_ORIGINS="http://localhost:3000"  # prod: your frontend origin(s)
# Comma-separated GitHub login or email allowlist for admin dashboard endpoints.
# Empty => nobody is an admin.
CONFIGURED_ADMINS=""                   # e.g. "alice,bob@my-org.com"
# URL of the LangGraph server the FastAPI side calls to trigger/stream runs.
# Defaults to http://localhost:2024 locally; set to your deployment URL in prod.
LANGGRAPH_URL="http://localhost:2024"

# === Linear (if using Linear trigger) ===
LINEAR_API_KEY=""                      # From step 5
LINEAR_WEBHOOK_SECRET=""               # From step 5

# === Slack (if using Slack trigger) ===
SLACK_BOT_TOKEN=""                     # From step 5
SLACK_BOT_USER_ID=""
SLACK_BOT_USERNAME=""
SLACK_SIGNING_SECRET=""
# Optional: Slack-specific default repo (falls back to DEFAULT_REPO_OWNER/NAME).
SLACK_REPO_OWNER=""
SLACK_REPO_NAME=""
# Optional: "Sign in with Slack" account linking (GitHub <-> Slack). See step 5.
SLACK_CLIENT_ID=""
SLACK_CLIENT_SECRET=""
SLACK_TEAM_ID=""                       # Optional; restrict linking to one workspace (T...)

# === Exa (optional — enables web search tool) ===
EXA_API_KEY=""                         # From https://dashboard.exa.ai

# === Reviewer / Analyzer (optional) ===
# LangSmith dataset where reviewer finding outcomes are recorded and read back by
# the analyzer. Defaults to "openswe-reviewer-outcomes" if unset.
REVIEWER_OUTCOMES_DATASET=""
# Single GitHub org whose members may trigger the agent on *public* repos.
# Empty => no public-repo gate (back-compat). Distinct from ALLOWED_GITHUB_ORGS.
PUBLIC_REPO_ORG_GATE=""

# === Sandbox (optional) ===
# Provider: langsmith (default), modal, daytona, runloop, or local. See CUSTOMIZATION.md.
SANDBOX_TYPE="langsmith"
DEFAULT_SANDBOX_SNAPSHOT_ID=""         # Required when SANDBOX_TYPE=langsmith (see step 4c)
DEFAULT_SANDBOX_SNAPSHOT_FS_CAPACITY_BYTES=""  # Root FS size in bytes (default: 32 GiB)
DEFAULT_SANDBOX_VCPUS=""               # vCPUs per sandbox (default: 4)
DEFAULT_SANDBOX_MEM_BYTES=""           # Memory in bytes per sandbox (default: 15 GiB)
DEFAULT_SANDBOX_IDLE_TTL_SECONDS=""    # Auto-stop after N seconds idle (default: 7200; 0 disables)
DEFAULT_SANDBOX_DELETE_AFTER_STOP_SECONDS=""  # Delete N seconds after stop (default: 86400; 0 disables)

# === Token Encryption ===
TOKEN_ENCRYPTION_KEY=""                # Generate with: openssl rand -base64 32
                                       # Supports key rotation: see "Rotating TOKEN_ENCRYPTION_KEY" below

Rotating TOKEN_ENCRYPTION_KEY

TOKEN_ENCRYPTION_KEY accepts either a single Fernet key or a comma- or newline-separated ordered list of keys, most-recent-first. New writes always encrypt under the first key; reads try every key in order. To rotate without invalidating already-stored GitHub tokens:

  1. Generate a new key: openssl rand -base64 32.
  2. Prepend it to TOKEN_ENCRYPTION_KEY, keeping the old key second:
    TOKEN_ENCRYPTION_KEY="<new_key>,<old_key>"
    
    Restart the server. New encryptions use <new_key>; existing ciphertexts still decrypt against <old_key>.
  3. Let active threads cycle (each fresh OAuth flow re-encrypts under the new key). After every active thread has re-authed, drop the old key:
    TOKEN_ENCRYPTION_KEY="<new_key>"
    
    Any thread still holding ciphertext under <old_key> will fail to decrypt and the user will be re-prompted to authenticate — same UX as if the thread had never authed.

7. Start the backend

Make sure ngrok is still running from step 2, then start the backend in a second terminal:

make dev          # uv run langgraph dev
# or: uv run langgraph dev --no-browser

langgraph dev serves all three graphs (agent, reviewer, analyzer) and the FastAPI app (agent.webapp:app) together on http://localhost:2024. The FastAPI app owns both the webhooks and the dashboard API:

Endpoint Purpose
POST /webhooks/github GitHub issue/PR/comment webhooks
POST /webhooks/linear Linear comment webhooks
GET /webhooks/linear Linear webhook verification
POST /webhooks/slack Slack event webhooks
POST /webhooks/slack/interactivity Slack Block Kit button interactions
GET /webhooks/slack Slack webhook verification
GET /dashboard/api/auth/login Dashboard GitHub OAuth login
GET /dashboard/api/auth/callback Dashboard GitHub OAuth callback (registered on the App in step 3b)
GET /dashboard/api/* Dashboard API (profiles, team settings, repos, review styles, threads, …)
GET /health Health check

make run (uvicorn agent.webapp:app --port 8000) serves the FastAPI app without the LangGraph runtime, on port 8000. The dashboard's Agents chat features call LangGraph, so for full local dev use make dev on :2024, not make run.

8. Run the dashboard (optional)

The dashboard is the web app in ui/. It's a static TanStack Start client that calls the FastAPI dashboard API from step 7. Run it in a third terminal:

cd ui
pnpm install
cat > .env <<'EOF'
VITE_DASHBOARD_API_BASE_URL="http://localhost:2024"
EOF
pnpm run dev          # vite dev --port 3000 -> http://localhost:3000

The dashboard needs VITE_DASHBOARD_API_BASE_URL in ui/.env pointing at the backend for local dev. The file is intentionally untracked because .env* files are gitignored.

The client calls ${VITE_DASHBOARD_API_BASE_URL}/dashboard/api/* with credentials: "include", so the backend's osw_session cookie rides along. Because the UI (:3000) and API (:2024) are different origins, the backend needs CORS enabled for the UI origin — set DASHBOARD_ALLOWED_ORIGINS="http://localhost:3000" (CORS is off unless this is set). Keep DASHBOARD_API_BASE_URL on an http:// URL locally so the cookie uses SameSite=Lax rather than Secure.

For the dashboard login to succeed, you need (from steps 3c / 6): GITHUB_APP_CLIENT_ID, GITHUB_APP_CLIENT_SECRET, DASHBOARD_JWT_SECRET, DASHBOARD_API_BASE_URL, DASHBOARD_BASE_URL, and DASHBOARD_ALLOWED_ORIGINS. To reach the admin pages (user mappings, etc.), add your GitHub login or email to CONFIGURED_ADMINS.

Other UI scripts: pnpm run build, pnpm run typecheck, pnpm run lint, pnpm run test.

9. Verify it works

GitHub

  1. Go to any issue in a repository where the app is installed
  2. Create or comment on an issue with: @openswe what files are in this repo?
  3. You should see:
    • A 👀 reaction on your comment within a few seconds
    • A new run in your LangSmith project
    • The agent replies with a comment on the issue

Linear

  1. Go to any Linear issue in a team you configured in LINEAR_TEAM_TO_REPO
  2. Add a comment: @openswe what files are in this repo?
  3. You should see:
    • A 👀 reaction on your comment within a few seconds
    • A new run in your LangSmith project
    • The agent replies with a comment on the issue

Slack

  1. In any channel where the bot is invited, start a thread
  2. Mention the bot: @open-swe what's in the repo?
  3. You should see a reply in the thread with the agent's response.

Dashboard

  1. With the backend (step 7) and UI (step 8) both running, open http://localhost:3000
  2. Click Sign in with GitHub — you'll be sent through the GitHub OAuth flow and back to the dashboard
  3. You should land logged-in and be able to see your profile/settings. If your GitHub login or email is in CONFIGURED_ADMINS, the Admin pages (e.g. User mappings) are available.

10. Production deployment

Production runs the backend and dashboard separately.

Backend — deploy on LangGraph Cloud / Platform:

  1. Push your code to a GitHub repository
  2. Connect the repo to LangGraph Cloud
  3. Set all environment variables from step 6 in the deployment config. Set DASHBOARD_BASE_URL and LANGGRAPH_URL to your production URLs (all https://). Set DASHBOARD_API_BASE_URL to the URL browsers use for dashboard API requests and OAuth callbacks: either the backend URL for direct cross-origin calls, or the dashboard/Vercel URL when a same-origin rewrite proxies /dashboard/api/*.
  4. Update your webhook URLs (Linear, Slack, GitHub App) and the GitHub App / Slack OAuth callback URLs to your production URLs (replace the ngrok / localhost values). The dashboard GitHub App callback must be <DASHBOARD_API_BASE_URL>/dashboard/api/auth/callback.

The langgraph.json at the project root defines the graphs and the custom HTTP app:

{
  "graphs": {
    "agent": "agent.server:traced_agent",
    "reviewer": "agent.reviewer:traced_reviewer_agent",
    "analyzer": "agent.analyzer:traced_analyzer",
    "chat": "agent.chat:traced_chat_agent",
    "scheduler": "agent.scheduler:get_scheduler",
    "ci_monitor": "agent.ci_monitor:get_ci_monitor"
  },
  "http": {
    "app": "agent.webapp:app"
  }
}

Dashboard — the ui/ app deploys to Vercel. The recommended production setup uses same-origin requests to /dashboard/api/* (leave VITE_DASHBOARD_API_BASE_URL empty), and ui/vercel.json rewrites those to the hosted LangGraph deployment. In this mode, set both DASHBOARD_API_BASE_URL and the GitHub App dashboard callback URL to the Vercel/dashboard origin (for example, https://your-dashboard.vercel.app/dashboard/api/auth/callback). The OAuth callback response then sets the osw_session cookie on the dashboard host, and later same-origin /dashboard/api/* requests include it. Update the rewrite destination in ui/vercel.json to your own LangGraph deployment URL.

Alternatively, you can run the dashboard as a direct cross-origin client: set VITE_DASHBOARD_API_BASE_URL to the hosted backend origin, set DASHBOARD_API_BASE_URL to that same backend origin, and include the dashboard origin in DASHBOARD_ALLOWED_ORIGINS.

Managed LangGraph Cloud + Vercel — the production env triad

The production runtime is managed LangGraph Cloud (backend) + Vercel (UI), not a self-hosted langgraph dev process. Get these three right, in the recommended same-origin setup:

Variable Where Value Why
LANGGRAPH_URL backend (LangGraph Cloud) the deployment's own https://…us.langgraph.app URL The server-side langgraph_client() (thread sidebar + run creation) connects here. It defaults to http://localhost:2024 for local dev; on managed it must be set or every dashboard/thread call ConnectErrors.
DASHBOARD_BASE_URL backend the dashboard origin, https://… Where token-free settings links point.
DASHBOARD_API_BASE_URL backend the same dashboard/Vercel origin, https://… Builds the OAuth redirect_uri and drives the session-cookie Secure; SameSite=None flags. Must include the https:// scheme — a schemeless value yields a schemeless redirect_uri and GitHub rejects login with "redirect_uri not associated with this application." (The server now coerces a schemeless value to https:// as a backstop, but set it explicitly.)
VITE_DASHBOARD_API_BASE_URL UI build (Vercel) empty Leave empty so the SPA calls relative /dashboard/api/* same-origin; ui/vercel.json rewrites those to the backend so the osw_session cookie rides along.

ui/vercel.json's rewrite destination must be the stable LangGraph deployment URL for the target environment (e.g. https://<deployment>-<hash>.us.langgraph.app) — a per-deployment alias that is stable across revisions. Never point it at an immutable per-deploy / preview URL. Vercel does not interpolate env vars into vercel.json rewrites, so this hostname is committed per Vercel project; set the dev project's rewrite to the dev deployment and the prod project's to the prod deployment.

Always start dashboard login on the stable alias / custom domain, not on an immutable per-deploy Vercel URL. The OAuth state cookie (osw_oauth_state) is intentionally host-only (a Domain cookie is unsafe across *.vercel.app, which is on the public-suffix list). If login starts on one host and GitHub redirects back to another, the cookie is dropped and the callback fails with "oauth state mismatch."

On managed the backend runs as multiple replicas. In-process caches that must stay coherent across replicas (notably the GitHub⇄work-email user mapping) are refreshed from the durable Store on the webhook hot paths (Slack and GitHub) before they are read.

Troubleshooting

Webhook not receiving events

  • Verify ngrok is running and the URL matches what's configured in GitHub/Linear/Slack
  • Check the ngrok web inspector at http://localhost:4040 for incoming requests
  • Ensure you enabled the correct event types (Comments → Create for Linear, app_mention for Slack, Issues + Issue comment for GitHub)
  • Webhook secrets are required — if GITHUB_WEBHOOK_SECRET, LINEAR_WEBHOOK_SECRET, or SLACK_SIGNING_SECRET is not set, all requests to that endpoint will be rejected with 401

GitHub authentication errors

  • Verify GITHUB_APP_ID, GITHUB_APP_PRIVATE_KEY, and GITHUB_APP_INSTALLATION_ID are set correctly
  • Ensure the GitHub App is installed on the target repositories
  • Check that the private key includes the full -----BEGIN RSA PRIVATE KEY----- and -----END RSA PRIVATE KEY----- lines

Dashboard login fails or won't stay logged in

  • 500 GITHUB_APP_CLIENT_ID not configured (or client secret): set GITHUB_APP_CLIENT_ID / GITHUB_APP_CLIENT_SECRET (step 3c) and DASHBOARD_JWT_SECRET.
  • OAuth redirect_uri mismatch / "redirect_uri not associated with this application": the GitHub App must list <DASHBOARD_API_BASE_URL>/dashboard/api/auth/callback as a callback URL (step 3b). Locally that's http://localhost:2024/dashboard/api/auth/callback. On managed, a common cause is a schemeless DASHBOARD_API_BASE_URL (e.g. dash.example.com instead of https://dash.example.com), which produces a schemeless redirect_uri; always set the full https:// URL.
  • "oauth state mismatch" after the GitHub redirect: login was started on a different host than the one GitHub redirected back to (typically an immutable per-deploy Vercel URL vs the stable alias). The osw_oauth_state cookie is host-only by design, so it isn't presented to the other host. Always start login on the stable alias / custom domain.
  • Login redirects but the session doesn't stick: this is almost always a cookie problem. Locally, keep DASHBOARD_API_BASE_URL on http:// (so cookies are SameSite=Lax); in prod use https:// for both API and frontend and add the frontend origin to DASHBOARD_ALLOWED_ORIGINS.
  • Login rejected with an org error: ALLOWED_GITHUB_ORGS gates dashboard login (and requires the App's Organization → Members: Read-only permission). See step 5.
  • Admin pages 403: add your GitHub login or email to CONFIGURED_ADMINS.

Dashboard UI can't reach the backend

  • Confirm the backend is running via make dev on :2024 (not make run on :8000).
  • Confirm ui/.env has VITE_DASHBOARD_API_BASE_URL=http://localhost:2024. If it's empty, the UI falls back to relative /dashboard/api/*, which only works behind the Vercel rewrite, not in local dev.

Sandbox creation failures

  • Verify LANGSMITH_API_KEY_PROD is set and valid
  • Check LangSmith sandbox quotas in your workspace settings
  • If the server refuses to start with DEFAULT_SANDBOX_SNAPSHOT_ID must be set, build a snapshot (see step 4c) and export its UUID
  • If you see Failed to create sandbox from snapshot '<id>', confirm the snapshot exists in your workspace and has status ready
  • If you get a 403 Forbidden error on the sandbox endpoints, your LangSmith workspace may not have sandbox access enabled — contact LangSmith support

Agent not responding to comments

  • For GitHub: ensure the comment or issue contains @openswe (case-insensitive), and the commenter has a user mapping (Admin → User mappings; see "Configure triggering surfaces"). Add any missing user with Add / update in that section.
  • For Linear: ensure the comment contains @openswe (case-insensitive)
  • For Slack: ensure the bot is invited to the channel and the message is an @mention
  • Check server logs for webhook processing errors

Token encryption errors

  • Ensure TOKEN_ENCRYPTION_KEY is set (generate with openssl rand -base64 32)
  • The key must be a valid 32-byte Fernet-compatible base64 string
  • For key rotation, TOKEN_ENCRYPTION_KEY may be a comma- or newline-separated list of keys (most-recent-first). See "Rotating TOKEN_ENCRYPTION_KEY" above.