adam/open-swe
1
0
Fork 0
mirror of https://github.com/Sea-Haven-Industries/open-swe.git synced 2026-10-01 09:43:14 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
1013 commits 12 branches 0 tags 41 MiB
Commit graph

3 commits

Author SHA1 Message Date
Johannes du Plessis
209132d355
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>
 2026-06-26 13:48:38 -07:00
Johannes du Plessis
4a55145bb1
feat: outcomes dataset + bootstrap/continual split via skills (#1365) 
* fix: reset stale sandbox creation sentinel

Co-authored-by: Johannes du Plessis <51395795+johannes117@users.noreply.github.com>

* fix: treat SANDBOX_CREATING as a timestamped cross-process lock

Only reset the sentinel when proven stale (older than the creation
timeout); otherwise wait for the worker that holds the lock so a
concurrent run does not create a duplicate sandbox.

* feat(analyzer): outcomes dataset + bootstrap/continual split via skills

Rename the review_style_analyzer graph to `analyzer` and split it into two
modes, plus capture reviewer finding outcomes for continual learning.

- Outcomes dataset: upsert resolved-by-commit (positive), dismissed (false
  positive), and GitHub/Slack thumbs findings into a single LangSmith dataset
  (openswe-reviewer-outcomes), keyed deterministically per finding+source.
  Emit points wired into update_finding, resolve_finding_thread, and the
  GitHub/Slack reaction handlers.
- Two playbooks delivered as deepagents skills (bootstrap-repo-analysis,
  continual-learning), served as virtual files via a CompositeBackend /skills/
  route + StateBackend (seeded into the run files channel at invoke time, never
  written to the sandbox). Mode is set by the launcher; continual runs fall
  back to the GitHub App installation token.
- Split launcher into start_bootstrap_analysis + start_continual_run; register
  a per-repo nightly continual-learning cron when bootstrap completes.
- New read_finding_outcomes tool feeds confirmed/dismissed findings back to the
  continual playbook.

Tests for outcome label mapping, skills helper, and cron idempotency.

* fix(analyzer): anchor continual cron runs to a real thread_id

The nightly continual-learning cron is threadless, and get_analyzer
early-returns an empty agent when configurable.thread_id is missing — so
every cron-launched run no-op'd before reading outcomes or saving a refined
prompt. Include the repo's deterministic analyzer thread_id in the continual
run configurable so the run executes; the threadless run carries no message
history, so nightly runs don't accumulate context.

* refactor(analyzer): move cron lifecycle calls out of the review-styles store

Drop the inline `analyzer_cron` imports from review_styles.py (added only to
dodge a circular import) by relocating the cron-trigger calls to the layer
above the store: registration to the save_review_style tool (after a prompt is
saved) and removal to the dashboard delete route. review_styles.py is now a
pure store again with top-level imports only.

* refactor: hoist reviewer_outcomes imports to module level

Move the two inline emit_finding_status_outcome imports introduced in this PR
(update_finding, resolve_finding_thread) to top-level imports. reviewer_outcomes
only depends on langsmith, so there is no circular import to avoid.

---------

Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
 2026-06-01 13:25:12 -07:00
Johannes du Plessis
82852f9eda
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 18:35:00 +00:00