feat(feedback): thumbs up/down on answers (MCP + chat) with an answer_log table - #1162

openRun now checks res.ok and does not populate the mapping editor
from error JSON. Successful target add/start/delete clears a stale
error banner.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Skip blank standard names so impact no longer calls get_nodes(""),
return 400 JSON instead of a 500 HTML page on graph failures, and
show impact warnings in the admin review UI.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Drop SQLite file-path probing for the agent index. Config help and
admin status now expect a Postgres URL and ignore other values.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Return JSON error bodies instead of HTML abort pages, ignore bad
repos.yaml on seed, and cover blank/duplicate/disabled target and
mapping edge cases in API and Admin UI tests.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Save writes the packaged harvester yaml. Start one-off uses a temp
copy so custom yaml does not overwrite it. Import review source is a
custom name or repos.yaml:<sha256-12>. GitHub org expand fills
repositories (skipping forks, archived, and existing owner/repo).

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Allowlist GitHub owner logins, quote the API path, cap yaml at 256KiB
before safe_load, and write packaged repos.yaml atomically.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

The 256KiB parse limit rejected a full GitHub org dump. Keep a
size cap, but large enough for up to 1000 repository entries.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Config is a sync write of github.com/org/ (or a single repo URL).
The indexer lists org repos at harvest time and classifies each as
OpenCRE or OWASP-agent material. Add org no longer calls GitHub.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Drop the dedicated Admin tab. Resources lists a virtual owasp-agent
row with status in spec; Start now runs the same harvest as oie_repo.
The built-in id is reserved, skipped on yaml seed, and cannot be
removed unless a leftover DB row exists.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Save, Add org, and one-off ingest probe api.github.com for each sources
URL before writing or starting harvest. Missing, private, or timed-out
orgs/repos return 400 with the canonical URL. The indexer still expands
reachable orgs later.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Admin home lists running jobs and links through to pipeline logs.
Config edits targets.yaml (cron, add-org, add-target form) plus every
.env.example env var with secrets redacted. Pipeline New can harvest
the packaged yaml or a one-off repository. Graph management reviews
per-link approve/deny/relink for each run. MyOpenCRE stays a separate
route.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Add org now writes packaged repos.yaml after the GitHub probe, matching
Add target. Graph management notes that link decisions stay on the staged
run until CRE-edge apply lands. Failed applies no longer look complete.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Remove the separate agent DB env var. Admin status and config read
DATABASE_URL / DEV_DATABASE_URL (same Postgres as CRE + pgvector).

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Admin invoked python /abs/scripts/run_oie_pipeline.py, so sys.path[0]
was scripts/ and `import application` failed. Other threads set
PYTHONPATH= at the repo root. Put the repo on sys.path in the script
and in the admin subprocess env.

IndexStore now defaults to DATABASE_URL / DEV_DATABASE_URL instead of
tmp/owasp_agent.sqlite. OWASP_AGENT_DB remains a test override.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

SQLAlchemy 2.1 maps postgresql:// to psycopg v3; pin psycopg2 via
postgresql+psycopg2://. Admin invoke_oie_cli uses the app Postgres URL
and refuses sqlite. `make admin-local` starts cre-postgres when the URL
is down, migrates, create_all, and launches Flask without CRE_CACHE_FILE.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Production and admin/agent now share main_db_env_keys (DATABASE_URL
before PROD_DATABASE_URL). Rewrite postgresql+psycopg:// to
+psycopg2. Gate local create_all behind ADMIN_LOCAL_CREATE_ALL.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Live defaults sync repos and persist to Postgres; auto-skip B/C without
LLM keys; async API with wait=true for sync probes. Treat harvester
degraded as non-fatal and surface chunks/files in stage detail.
Pass --max-repos through the CLI so admin caps stay effective.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Empty locale stubs in Top10 aborted the whole repo harvest; skip and
continue so golden-set ingest can finish. Admin agent status now counts
all public tables and reports MAX(fetched_at) as last_sync.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Map only Module C safety-path degraded as non-fatal so A/B degraded
halts before B/C again. Reject max_repos < 1. Always skip B/C unless
the client opts in so LLM keys cannot launch expensive stages from New.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Admin /rest JSON errors no longer collapse to HTML "Resource Not found".
List backfills missing staged_change_set rows. Graph UI shows Accept
feedback, disables dead actions, and explains empty CRE changesets on
OIE harvest runs (ops live in harvest_input).

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

Orchestrator JSON now includes engine, flags, graph_path, and visited.
Skip stage details explain caller flags; harvester detail notes HEAD
checkpoints when chunks=0. Admin records oie_trace and richer started JSON.

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

OIE harvest fills harvest_input but not the explorer graph; run
cre.py --upstream_sync before Flask unless the cre/node tables
already have data (SKIP_UPSTREAM_SYNC / FORCE_UPSTREAM_SYNC overrides).

Co-authored-by: Spyros <northdpole@users.noreply.github.com>

…ed Google email

auth_callback matched LOGIN_ALLOWED_DOMAINS with email.endswith(domain), so
attacker@evilowasp.org passed for owasp.org, and never read email_verified.
Reuse the roles domain rule (exact domain or subdomain, same as the MCP
sign-in) and refuse unverified addresses, named admins included.
Existing callback tests now carry email_verified.

…base

* trim/1127:
  test(owasp-agent): sync and concepts must not import application.database

* trim/1137:
  test(owasp-agent): sync and concepts must not import application.database

* trim/1136:
  test(owasp-agent): sync and concepts must not import application.database

…and counts real vectors

* trim/1142:
  fix(admin): CRE embedding guard only treats a missing table as empty and counts real vectors
  test(owasp-agent): sync and concepts must not import application.database

* trim/1155:
  fix(admin): CRE embedding guard only treats a missing table as empty and counts real vectors
  test(owasp-agent): sync and concepts must not import application.database

* trim/1159:
  fix(admin): CRE embedding guard only treats a missing table as empty and counts real vectors
  test(owasp-agent): sync and concepts must not import application.database

* trim/1158:
  fix(admin): CRE embedding guard only treats a missing table as empty and counts real vectors
  test(owasp-agent): sync and concepts must not import application.database

…-style domains, require email_verified is True

…d of patch.dict in the embed_sync tests

* trim/1127:
  test(owasp-agent): restore the environment with a plain helper instead of patch.dict in the embed_sync tests

…embed_fn instead of patching embed_sync

* trim/1137:
  test(owasp-agent): exercise the meta-embedding push through injected embed_fn instead of patching embed_sync
  test(owasp-agent): restore the environment with a plain helper instead of patch.dict in the embed_sync tests

# Conflicts:
#	application/tests/oie_scheduler/runner_test.py
#	application/utils/oie_scheduler/jobs.py

…he new opt-in assertions

* trim/1136:
  test(oie-scheduler): restore the environment with a plain helper in the new opt-in assertions
  test(owasp-agent): exercise the meta-embedding push through injected embed_fn instead of patching embed_sync
  test(owasp-agent): restore the environment with a plain helper instead of patch.dict in the embed_sync tests

# Conflicts:
#	application/tests/admin_panel_test.py
#	application/utils/admin_panel/service.py

…vironment with a plain helper in the #1142 tests

* trim/1142:
  docs(staging): use placeholder tester addresses; test: restore the environment with a plain helper in the #1142 tests
  test(oie-scheduler): restore the environment with a plain helper in the new opt-in assertions
  test(owasp-agent): exercise the meta-embedding push through injected embed_fn instead of patching embed_sync
  test(owasp-agent): restore the environment with a plain helper instead of patch.dict in the embed_sync tests

* trim/1155:
  docs(staging): use placeholder tester addresses; test: restore the environment with a plain helper in the #1142 tests
  test(oie-scheduler): restore the environment with a plain helper in the new opt-in assertions
  test(owasp-agent): exercise the meta-embedding push through injected embed_fn instead of patching embed_sync
  test(owasp-agent): restore the environment with a plain helper instead of patch.dict in the embed_sync tests

…ex_routing_test

* trim/1159:
  test(owasp-agent): restore the environment with a plain helper in index_routing_test
  docs(staging): use placeholder tester addresses; test: restore the environment with a plain helper in the #1142 tests
  test(oie-scheduler): restore the environment with a plain helper in the new opt-in assertions
  test(owasp-agent): exercise the meta-embedding push through injected embed_fn instead of patching embed_sync
  test(owasp-agent): restore the environment with a plain helper instead of patch.dict in the embed_sync tests

* trim/1158:
  test(owasp-agent): restore the environment with a plain helper in index_routing_test
  docs(staging): use placeholder tester addresses; test: restore the environment with a plain helper in the #1142 tests
  test(oie-scheduler): restore the environment with a plain helper in the new opt-in assertions
  test(owasp-agent): exercise the meta-embedding push through injected embed_fn instead of patching embed_sync
  test(owasp-agent): restore the environment with a plain helper instead of patch.dict in the embed_sync tests

… callback instead of patching them; plain env helper; placeholder addresses in .env.example