Repository navigation
feat(librarian): Module C retrieval upgrades (stack base) - #1088
Merged
Merged
Conversation
Contributor
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. 🗂️ Base branches to auto review: 1
Please check the settings in the CodeRabbit UI or the
 |
| Layer / File(s) | Summary |
|---|---|
OIE taxonomy and metadata workflows application/utils/librarian/oie_taxonomy.py, application/utils/librarian/metadata_enricher.py, scripts/oie_*, scripts/oie_taxonomy_bootstrap.py, application/tests/librarian/oie_*, application/tests/librarian/metadata_enricher_test.py |
Adds OIE taxonomy indexing, metadata tagging and enrichment, and scripts for preparing and checking metadata. |
Classification, CRE priors, and transfer indexes application/utils/librarian/problem_class.py, application/utils/librarian/standard_link_seed.py, application/utils/librarian/aix_llm_link_seed.py, application/utils/librarian/cre_prior.py, application/utils/librarian/edition_remap.py, application/utils/librarian/neighbor_transfer.py, application/utils/librarian/control_name_seed.py, application/utils/librarian/umbrella_promote.py, application/tests/librarian/*seed_test.py, application/tests/librarian/*prior_test.py, application/tests/librarian/*transfer_test.py, application/tests/librarian/problem_class_test.py, application/tests/librarian/edition_remap_test.py, application/tests/librarian/umbrella_promote_test.py |
Adds classification from catalog IDs and text, then uses standard links, taxonomy priors, edition remaps, neighboring standards, names, and parent relationships to derive preferred CRE IDs and retrieval constraints. |
Retrieval and ranking composition application/utils/librarian/candidate_retriever.py, application/utils/librarian/dual_index_retriever.py, application/utils/librarian/hub_leaf_cage.py, application/utils/librarian/standard_hop_retriever.py, application/utils/librarian/cross_encoder.py, application/utils/librarian/title_boost.py, application/utils/librarian/margin_cutoff.py, application/utils/librarian/focus_query.py, application/utils/librarian/config_loader.py, application/tests/librarian/candidate_retriever_test.py, application/tests/librarian/dual_index_retriever_test.py, application/tests/librarian/hub_leaf_cage_test.py, application/tests/librarian/standard_hop_retriever_test.py, application/tests/librarian/cross_encoder_test.py, application/tests/librarian/title_boost_test.py, application/tests/librarian/margin_cutoff_test.py, application/tests/librarian/focus_query_test.py, application/tests/librarian/config_loader_test.py |
Adds allowlisted retrieval, name/body dual indexing, hub/leaf cages, Standard-to-CRE hops, and hybrid ranking controls. Configuration supplies the related retrieval and scoring settings. |
Pipeline decisions and safety handling application/utils/librarian/pipeline.py, application/utils/librarian/decision_engine.py, application/utils/librarian/explicit_link_resolver.py, application/utils/librarian/cre_gap_suggester.py, application/utils/librarian/cre_registry.py, application/utils/librarian/shortlist_judge.py, application/utils/librarian/llm_safety_guard.py, application/utils/librarian/emitter.py, application/utils/librarian/queue_runner.py, application/utils/librarian/schemas.py, application/tests/librarian/pipeline_test.py, application/tests/librarian/decision_engine_test.py, application/tests/librarian/explicit_link_resolver_test.py, application/tests/librarian/cre_gap_suggester_test.py, application/tests/librarian/cre_registry_test.py, application/tests/librarian/shortlist_judge_test.py, application/tests/librarian/llm_safety_guard_test.py, application/tests/librarian/queue_runner_test.py, application/tests/librarian/fixtures/golden_dataset.schema.json |
Adds authoritative OpenCRE URL handling, shortlist judging, CRE-gap proposals, membership grounding, and LLM safety verdicts. Review decisions can include two candidate IDs, and the emitter formats gap proposals as suggested links. |
CRE text, summaries, and component wiring application/utils/librarian/cre_text.py, application/utils/librarian/cre_summary.py, application/utils/librarian/factory.py, application/utils/librarian/README.md, scripts/oie_generate_cre_summaries.py, scripts/evaluate_librarian.py, application/tests/librarian/cre_text_test.py, application/tests/librarian/cre_summary_test.py, application/tests/librarian/factory_test.py |
Adds linked-Standard text enrichment and cached CRE summaries, with optional summary vectors and dual-index retrieval. Component construction and evaluation wire in the configured retrieval, text, and reranking options. |
| Embedding acquisition and retry handling |
|---|
application/prompt_client/prompt_client.py, application/prompt_client/llm_error_utils.py, application/prompt_client/litellm_router.py, application/utils/librarian/embedding_quality.py, scripts/audit_standard_embeddings.py, scripts/oie_sync_cre_hub.py, application/tests/prompt_client_incremental_embeddings_test.py, application/tests/prompt_client_plain_text_fetch_test.py, application/tests/librarian/embedding_quality_test.py, application/tests/litellm_router_test.py |
| Layer / File(s) | Summary |
|---|---|
Embedding text and vector handling application/prompt_client/prompt_client.py, application/prompt_client/llm_error_utils.py, application/prompt_client/litellm_router.py, application/utils/librarian/embedding_quality.py, scripts/audit_standard_embeddings.py, scripts/oie_sync_cre_hub.py, application/tests/prompt_client_incremental_embeddings_test.py, application/tests/prompt_client_plain_text_fetch_test.py, application/tests/librarian/embedding_quality_test.py, application/tests/litellm_router_test.py |
Adds GitHub raw-URL selection, bounded plain-text fetching, content filtering, selected database-field fallback text, and an optional disk cache for single-string vectors. LiteLLM retries transient connection and timeout errors as well as rate-limit errors. Adds embedding-quality reporting and CRE embedding synchronization scripts. |
Estimated code review effort: 5 (Critical) | ~100 minutes
Change: Feature
Possibly related PRs
- OWASP/OpenCRE#1172: This PR changes the Librarian decision results and pipeline constructor used by the mapping workflow in
#1172.
Merge Risk: 
44dc4
-
Several retrieval and safety paths can still produce incorrect links or bypass intended safeguards. Resolve the material issues before merging; the two previously reported test failures no longer appear to apply.
Security architecture risk: dd480
-
The new database-copy command can expose credentials and does not reliably enforce its intended connection boundaries. It requires operator execution and database permissions, limiting reachability, but can affect the destination’s shared catalog.
Retained concerns
- Medium · security · observed: The newly added sync command enforces its local-destination policy using only the URI authority hostname, while passing the complete URI to libpq. Query-string host or hostaddr parameters can select a remote destination without the acknowledgement flag. The resulting operation replaces every destination CRE embedding and upserts referenced CRE metadata. The base command had an analogous guard weakness, but only copied embeddings.
- High · security · inferred: The new PostgreSQL source path relies on caller or environment TLS settings except for selected URLs, where it adds sslmode=require. This does not itself guarantee server identity verification. Under permissive effective settings, a network-positioned attacker could alter copied catalog content. Permissive source connections predate this PR, but the new command also accepts source metadata for insertion or overwrite in the CRE catalog.
- High · security · observed: The new command prints the complete destination URL after committing its writes, including any supplied username and password. Anyone able to read captured stdout may obtain the supplied database identity. The impact is bounded by that identity’s actual grants, which are unknown. The base command already printed destination URLs, so this is an additional credential-output path rather than a novel exposure class.
Security review details
-
Security Blast Radius
- observed — The direct mutation scope is all CRE embeddings in the selected destination database and metadata for source-referenced CRE IDs. There is no tenant or ownership predicate in these writes. Execution requires a supplied database identity; exposure beyond these operations depends on its grants and is not established by the available deployment evidence.
Security Findings and Attack Paths
- inferred — The three retained findings cover distinct attack prerequisites: access to captured stdout for destination credentials, influence over a destination URI plus usable database authentication for the local-policy bypass, and network interception under permissive effective TLS settings for source tampering. None establishes unauthenticated web reachability or automatic cross-tenant access.
Trust Boundaries and Controls
- observed — Source database content crosses into destination catalog authority through parameterized upserts and inserts. The command rejects an empty usable embedding set before opening its destination transaction and defaults to a local destination. These controls protect against empty-source destruction and SQL value injection, but do not bind the effective libpq endpoint or authenticate the source server.
Resilience and Maintainability Implications
- observed — Destination metadata upsert and embedding replacement share one transaction, containing partial mutation on failure before commit. Counts and credential-bearing output occur after the explicit commit, so a subsequent reporting failure cannot undo the writes. There is no explicit synchronization between concurrent sync runs or built-in restoration of the previous catalog; rerunning identical input reconstructs the embedding set but leaves stale CRE records.
- observed — The shortlist disk cache is keyed by query text and ordered candidate IDs, without tenant, authorization, model-version, or run identity. Writes are not atomic; malformed reads become cache misses, and cached IDs are revalidated before use. Whether reuse across authorization contexts violates an ownership contract remains unresolved, so no tenant-isolation finding is asserted.
Hardening Proposals
- proposed — Redact connection credentials from operational output, enforce destination policy against libpq’s effective connection parameters, and require authenticated TLS for remote database connections. Use database roles limited to the required catalog operations to bound both leaked-credential and unintended-destination impact.
Pre-merge checks: 
-
Check name Status Explanation Title check The title "feat(librarian): Module C retrieval upgrades (stack base)" matches the main change. The changeset adds dual-index retrieval, CRE summaries, prior-cage retrieval, edition remapping, and rela… Description check The description states that the PR covers librarian retrieval: dual-index C.1 retrieval, CRE summaries, prior cage, hub seed, edition remap, factory defaults, and embedding text hygiene. These match t… Docstring Coverage Docstring coverage is inconclusive: files from earlier reviews still need to be rechecked within the check's analysis limits. Not blocking. Linked Issues check Check skipped because no linked issues were found for this pull request. Out of Scope Changes check Check skipped because no linked issues were found for this pull request. Full details: Title check
-
Explanation
The title "feat(librarian): Module C retrieval upgrades (stack base)" matches the main change. The changeset adds dual-index retrieval, CRE summaries, prior-cage retrieval, edition remapping, and related librarian tests and utilities. The title is specific to the librarian retrieval scope. The phrase "stack base" refers to the PR's position in a stack, which is accurate but not a description of the code change itself.
Full details: Description check
-
Explanation
The description states that the PR covers librarian retrieval: dual-index C.1 retrieval, CRE summaries, prior cage, hub seed, edition remap, factory defaults, and embedding text hygiene. These match the changeset. It also gives a test command, a scope check, and the stack order. It is related to the changeset.
-
Finishing Touches
-
Generate docstrings
-
- Commit to this branch
- Create a new PR
Generate unit tests (beta)
-
- Commit to this branch
- Create a new PR
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
This was referenced Sep 13, 2026
northdpole
force-pushed
the
wip/oie-pipeline-eval-harness
branch
from
September 13, 2026 17:50
f16272c to
c6bf04b
Compare
4 tasks
northdpole
changed the title
northdpole
force-pushed
the
wip/oie-pipeline-eval-harness
branch
from
September 22, 2026 15:57
9364f9b to
d346bf7
Compare
northdpole
added a commit
that referenced
this pull request
Sep 22, 2026
Unshaped dump for Cloud Agent to split into a clean commit stack on #1088.
cursor
Bot
force-pushed
the
wip/oie-pipeline-eval-harness
branch
from
September 22, 2026 23:09
59957ab to
81434d2
Compare
Collaborator
Author
Please stress-test GitHub ingestPoint-and-shoot ingest is on this branch: give it a public GitHub URL → harvest markdown → Module B noise filter → print knowledge. Try to break it. Prereq — Gemini key (required for real classification)Module B needs cp .env.example .env
# set: GEMINI_API_KEY=…Sanity check (prints python -c "import os; from pathlib import Path; from dotenv import load_dotenv; load_dotenv(Path('.env')); print('ok' if os.getenv('GEMINI_API_KEY') else 'missing')"Without a key, use What to rungit fetch origin && git checkout wip/oie-pipeline-eval-harness
# venv + usual local setup, then .env with GEMINI_API_KEY
make ingest URL=https://github.com/OWASP/ASVS
make ingest URL=https://github.com/OWASP/ASVS/tree/master/5.0/en
./scripts/ingest_github.sh https://github.com/OWASP/CheatSheetSeries
python cre.py --ingest_github https://github.com/OWASP/CheatSheetSeriesNo key on hand (last resort): ./scripts/ingest_github.sh https://github.com/OWASP/ASVS --ingest_keep_all
Good looks like
Please try to break itWeird URLs, huge/tiny repos, table-heavy standards (ASVS/AISVS), re-runs, non-GitHub hosts (should fail cleanly). How to reportComment here with: (1) exact command (2) repo URL (3) whether GEMINI was configured — yes/no, not the key (4) expected vs actual (stderr counts + short stdout sample) (5) keep-all or not. This is not the eval/accuracy scorer — just “does URL ingest dump useful knowledge?” |
Collaborator
Author
|
Hey everyone — we’re trying something new. Thanks to our awesome community, a ton of work from our GSoC students, and the maintainer team, we now have a point-and-shoot way to feed a public GitHub repo into OpenCRE’s OIE pipeline and see what it keeps as security knowledge. You can find it on the WIP PR: #1088 (branch This is version zero of the OWASP agent (ingest → classify → show knowledge). We’re fine-tuning toward v1 and prepping for a release around AppSec California / AppSec SF — and we want beta testers from the community to hammer it before then. How you can help
git fetch origin && git checkout wip/oie-pipeline-eval-harness
cp .env.example .env # if needed
# set GEMINI_API_KEY=… in .envSanity check (prints python -c "import os; from pathlib import Path; from dotenv import load_dotenv; load_dotenv(Path('.env')); print('ok' if os.getenv('GEMINI_API_KEY') else 'missing')"
make ingest URL=https://github.com/OWASP/ASVS
make ingest URL=https://github.com/OWASP/ASVS/tree/master/5.0/en
./scripts/ingest_github.sh https://github.com/OWASP/CheatSheetSeries(Make can’t take a bare No Gemini key? Last resort only (not a real classification test): ./scripts/ingest_github.sh https://github.com/OWASP/ASVS --ingest_keep_all
Please try weird URLs, huge/tiny repos, table-heavy standards (ASVS / AISVS), re-runs, non-GitHub hosts. This is not the accuracy/eval scorer — it’s “does URL ingest dump useful knowledge?” Thanks for beta-testing v0 with us. |
Collaborator
Author
Update: any LiteLLM provider / mid-size chat modelModule B already runs through LiteLLM. Ingest preflight now accepts whatever provider key matches your model — not Gemini-only. In OPENAI_API_KEY=…
CRE_NOISE_FILTER_LLM_MODEL=openai/gpt-4o-mini
# or ANTHROPIC_API_KEY=… + anthropic/claude-haiku-4-5
# or GEMINI_API_KEY=… + gemini/gemini-2.5-flash
# or Groq / Mistral / OpenRouter / … (see .env.example)Or per run: make ingest URL=https://github.com/OWASP/ASVS INGEST_ARGS='--ingest_model openai/gpt-4o-mini'
./scripts/ingest_github.sh https://github.com/OWASP/ASVS --ingest_model anthropic/claude-haiku-4-5Prefer a mid-size chat model (gpt-4o-mini, Claude Haiku/Sonnet, Gemini Flash, Llama 70B on Groq, …). Don’t paste keys in the PR. |
Collaborator
Author
|
Wording tweak for beta testers: any LiteLLM chat model works; mid-size models usually work best (cheap + reliable for the classify gate — tiny stubs can be flaky, frontier models are optional spend). |
northdpole
marked this pull request as ready for review
September 23, 2026 19:30
northdpole
requested review from
Pa04rth,
paoga87 and
robvanderveer
as code owners
September 23, 2026 19:30
Contributor
|
@northdpole Stress-testing Command / setup
Expected vs actual
Notes
|
northdpole
added a commit
that referenced
this pull request
Sep 28, 2026
…name-packer Sync with PR #1088 upstream. Prefer dual-index librarian defaults on conflicts; take 1088-only harness/full-pipeline/NIST/harvester additions.
northdpole
added a commit
that referenced
this pull request
Sep 29, 2026
Dual-index retrieval, CRE summaries, prior cage, hub seed, edition remap, and related librarian defaults. Rewrites #1088 as the reviewable base; later stacked PRs add harvester, B2 eval, fixtures, mappings, and full-pipeline docs.
northdpole
force-pushed
the
wip/oie-pipeline-eval-harness
branch
from
September 29, 2026 09:51
b0e5e65 to
9fb6d73
Compare
northdpole
changed the title
This was referenced Sep 29, 2026
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
This was referenced Sep 29, 2026
- Refuse plain-text fetch redirects that leave GitHub raw hosts (SSRF). - Re-check expected embedding dimension on query-cache hits. - Fill phrase-only oie blocks in apply_oie_fill_if_missing without losing phrases. - Expose last_preferred_cre_ids on StandardHopRetriever and honour its allowlist. - Commit each metadata enrichment so a later failure cannot roll back earlier work. - Load .env from the repo root in the enrichment scripts. - Parse hostnames for CRE hub sync SSL and loopback checks; batch SQLite IN lookups; stop echoing database credentials.
Wire LlmSafetyGuard (LiteLLM) as the live factory default so runs evaluate safety instead of NullSafetyGuard. Apply the guard on every decision path, prefer safety reason codes over empty shortlists, and cover a synthetic adversarial corpus in unit tests.
northdpole
added a commit
that referenced
this pull request
Oct 14, 2026
Optional requirement extract (off/auto/on) plus GitHub tarball/URL ingest wiring for Module B → C. Stacked on #1088 librarian base.
northdpole
force-pushed
the
wip/oie-pipeline-eval-harness
branch
from
October 14, 2026 08:48
6b95fa1 to
7fe303e
Compare
5 tasks done
northdpole
added a commit
that referenced
this pull request
Oct 21, 2026
Dual-index retrieval, CRE summaries, prior cage, hub seed, edition remap, and related librarian defaults. Rewrites #1088 as the reviewable base; later stacked PRs add harvester, B2 eval, fixtures, mappings, and full-pipeline docs.
northdpole
added a commit
that referenced
this pull request
Oct 21, 2026
Optional requirement extract (off/auto/on) plus GitHub tarball/URL ingest wiring for Module B → C. Stacked on #1088 librarian base.
|
Tested the ingest harness on branch stack/oie-p7-full-pipeline-docs Command: Actual: chunks are prefixed with "Standard: DocuQuery" and "Version: 2005". Cause: ChunkRecordBuilder._version_id (application/utils/harvester/chunk_record_builder.py) Control: ingesting https://github.com/jlevy/the-art-of-command-line the same Note: grepping application/tests for Fix ideas I'd suggest:
Happy to open a PR against the stack if you want one. |
Collaborator
Author
|
Review gate result for #1088 (stack base): TRIM. Scope: 91 files, +17.9k/-305. This is the Module C retrieval base. 24 of the 27 new modules and scripts are imported or wired later in the stack. There are no migrations, no new dependencies, and no secrets or hardcoded hosts. The new GitHub plain-text fetch parses the host and re-validates redirects, with spoof tests. Requested trims before merge:
To confirm before staging:
Follow-ups, not blocking: the default-on LLM safety guard sends chunk text to an external LLM; flag-gated losing experiments ( |
- Remove the unwired context_enrich experiment (module and test); it is preserved on the staging-experiments branch. - Drop oie_enrich_neighbor_phrases.py and oie_enrich_node_metadata_llm.py (no references anywhere in the stack). - Revert golden_dataset.json and build_golden_dataset.py to main. The 190 gold:asvs5 rows depend on the ASVS 5 fixture, which lands in the ASVS fixtures PR; they move there. The CRE_GAP schema enum stays. - Delete the duplicated _env_bool/_env_csv block in librarian config_loader.
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
…nto trim/1115 * origin/wip/oie-pipeline-eval-harness: chore(oie): trim #1088 for the staging rollout
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1119: chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout # Conflicts: # requirements-dev.txt
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1120: trim(#1120): fix merge-conflict markers, guard full pipeline, drop dead experiments style: black-format llm_safety_guard.py chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout # Conflicts: # application/utils/librarian/factory.py # docs/oie/full-pipeline.md # scripts/oie_owasp_eval/run_fullpipe_detached.sh
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1125: trim(#1120): fix merge-conflict markers, guard full pipeline, drop dead experiments style: black-format llm_safety_guard.py chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1127: trim(#1127): drop adversarial asker/batch scripts and canned CRE follow-ups, set source_channel in raw embedding writers trim(#1120): fix merge-conflict markers, guard full pipeline, drop dead experiments style: black-format llm_safety_guard.py chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout # Conflicts: # cre.py # scripts/owasp_agent_adversarial_asker.py # scripts/owasp_agent_adversarial_batch_run.py
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1137: trim(#1137): drop brevity-bot skills and stale handover doc, restore README/dev setup, loopback guard for run_admin_local.sh trim(#1127): drop adversarial asker/batch scripts and canned CRE follow-ups, set source_channel in raw embedding writers trim(#1120): fix merge-conflict markers, guard full pipeline, drop dead experiments style: black-format llm_safety_guard.py chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1136: trim(#1137): drop brevity-bot skills and stale handover doc, restore README/dev setup, loopback guard for run_admin_local.sh trim(#1127): drop adversarial asker/batch scripts and canned CRE follow-ups, set source_channel in raw embedding writers trim(#1120): fix merge-conflict markers, guard full pipeline, drop dead experiments style: black-format llm_safety_guard.py chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout test: embedding-failure scheduler test needs graph filing enabled now that it defaults off style: format files flagged by CI (black, prettier) Make event-repo harvesting opt-in (OIE_HARVEST_EVENTS) Make event page parsing linear-time and clean Playwright fallback text Embed auto-filed graph nodes in the CRE expansion job Parse OWASP event pages and harvest event repos # Conflicts: # application/utils/oie_scheduler/jobs.py # scripts/oie_owasp_eval/run_full_pipeline.py
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1142: trim(#1142): admin upstream-sync-and-ingest no longer replaces existing CRE embeddings trim(#1142): no credentialed CORS without an allowlist, OIE workers opt-in in import-all, guard levers-matrix queue reset trim(#1137): drop brevity-bot skills and stale handover doc, restore README/dev setup, loopback guard for run_admin_local.sh trim(#1127): drop adversarial asker/batch scripts and canned CRE follow-ups, set source_channel in raw embedding writers trim(#1120): fix merge-conflict markers, guard full pipeline, drop dead experiments style: black-format llm_safety_guard.py chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout test: embedding-failure scheduler test needs graph filing enabled now that it defaults off style: format files flagged by CI (black, prettier) Make event-repo harvesting opt-in (OIE_HARVEST_EVENTS) Make event page parsing linear-time and clean Playwright fallback text Embed auto-filed graph nodes in the CRE expansion job Parse OWASP event pages and harvest event repos # Conflicts: # scripts/owasp_agent_adversarial_asker.py # scripts/owasp_agent_adversarial_batch_run.py
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1155: trim(#1155): drop laptop-local edge probe scripts, make node metadata seeding opt-in, fix empty-key blurb match trim(#1142): admin upstream-sync-and-ingest no longer replaces existing CRE embeddings trim(#1142): no credentialed CORS without an allowlist, OIE workers opt-in in import-all, guard levers-matrix queue reset trim(#1137): drop brevity-bot skills and stale handover doc, restore README/dev setup, loopback guard for run_admin_local.sh trim(#1127): drop adversarial asker/batch scripts and canned CRE follow-ups, set source_channel in raw embedding writers trim(#1120): fix merge-conflict markers, guard full pipeline, drop dead experiments style: black-format llm_safety_guard.py chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout test: embedding-failure scheduler test needs graph filing enabled now that it defaults off style: format files flagged by CI (black, prettier) Make event-repo harvesting opt-in (OIE_HARVEST_EVENTS) Make event page parsing linear-time and clean Playwright fallback text Embed auto-filed graph nodes in the CRE expansion job Parse OWASP event pages and harvest event repos
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1159: trim(#1159): drop cambridge->boston alias, dead locals, test env/tempfile leaks trim(#1155): drop laptop-local edge probe scripts, make node metadata seeding opt-in, fix empty-key blurb match trim(#1142): admin upstream-sync-and-ingest no longer replaces existing CRE embeddings trim(#1142): no credentialed CORS without an allowlist, OIE workers opt-in in import-all, guard levers-matrix queue reset trim(#1137): drop brevity-bot skills and stale handover doc, restore README/dev setup, loopback guard for run_admin_local.sh trim(#1127): drop adversarial asker/batch scripts and canned CRE follow-ups, set source_channel in raw embedding writers trim(#1120): fix merge-conflict markers, guard full pipeline, drop dead experiments style: black-format llm_safety_guard.py chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout test: embedding-failure scheduler test needs graph filing enabled now that it defaults off style: format files flagged by CI (black, prettier) Make event-repo harvesting opt-in (OIE_HARVEST_EVENTS) Make event page parsing linear-time and clean Playwright fallback text Embed auto-filed graph nodes in the CRE expansion job Parse OWASP event pages and harvest event repos
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1158: trim(#1158): strip_injection no longer truncates legitimate questions; correct default-off claims trim(#1159): drop cambridge->boston alias, dead locals, test env/tempfile leaks trim(#1155): drop laptop-local edge probe scripts, make node metadata seeding opt-in, fix empty-key blurb match trim(#1142): admin upstream-sync-and-ingest no longer replaces existing CRE embeddings trim(#1142): no credentialed CORS without an allowlist, OIE workers opt-in in import-all, guard levers-matrix queue reset trim(#1137): drop brevity-bot skills and stale handover doc, restore README/dev setup, loopback guard for run_admin_local.sh trim(#1127): drop adversarial asker/batch scripts and canned CRE follow-ups, set source_channel in raw embedding writers trim(#1120): fix merge-conflict markers, guard full pipeline, drop dead experiments style: black-format llm_safety_guard.py chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout test: embedding-failure scheduler test needs graph filing enabled now that it defaults off style: format files flagged by CI (black, prettier) Make event-repo harvesting opt-in (OIE_HARVEST_EVENTS) Make event page parsing linear-time and clean Playwright fallback text Embed auto-filed graph nodes in the CRE expansion job Parse OWASP event pages and harvest event repos # Conflicts: # application/mcp/ask.py # application/web/web_main.py
northdpole
added a commit
that referenced
this pull request
Nov 1, 2026
* trim/1162: (27 commits) docs(env): document CHAT_BLOCK_UNVERIFIED (default on, changes public chat answers) trim(#1162): drop prune_noise_chunks (unguarded DELETE on embeddings), validate authority weights before use, document ANSWER_LOG_*, no host paths in sync_check_staging trim(#1158): strip_injection no longer truncates legitimate questions; correct default-off claims trim(#1159): drop cambridge->boston alias, dead locals, test env/tempfile leaks trim(#1155): drop laptop-local edge probe scripts, make node metadata seeding opt-in, fix empty-key blurb match trim(#1142): admin upstream-sync-and-ingest no longer replaces existing CRE embeddings trim(#1142): no credentialed CORS without an allowlist, OIE workers opt-in in import-all, guard levers-matrix queue reset trim(#1137): drop brevity-bot skills and stale handover doc, restore README/dev setup, loopback guard for run_admin_local.sh trim(#1127): drop adversarial asker/batch scripts and canned CRE follow-ups, set source_channel in raw embedding writers trim(#1120): fix merge-conflict markers, guard full pipeline, drop dead experiments style: black-format llm_safety_guard.py chore(oie): trim #1118 for the staging rollout chore(oie): trim #1117 for the staging rollout chore(oie): trim #1116 and guard destructive eval scripts chore(harvester): trim #1115 for the staging rollout chore(oie): trim #1088 for the staging rollout fix(auth): harden web sign-in found by adversarial review fix(auth): /auth/user needs any role, not admin fix(auth): carry the PKCE code_verifier from login to callback fix(auth): build a fresh OAuth flow per login instead of a shared singleton ...
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Stack base for OIE Module C. Rewritten from the monolithic tip so reviews stay scoped.
LiteLLM shared router already landed via #1102 — this slice is librarian retrieval only:
Full tip backup (pre-split): commit
b0e5e65b/ local refrefs/backup/oie-1088-full.Stack (merge bottom → top)
Test plan
python -m unittest discover -s application/tests/librarian -p '*_test.py'scripts/oie_owasp_eval/fixtures/b2_sources)Stack (merge bottom → top)
wip/oie-pipeline-eval-harness→main)stack/oie-p2-harvester→ feat(librarian): Module C retrieval upgrades (stack base) #1088)stack/oie-p3-b2-harness→ feat(harvester): requirement extract and GitHub URL ingest #1115)stack/oie-p4-asvs-fixtures→ feat(oie): B2 eval harness and CRE neighborhood scorer #1116)stack/oie-p5-nist-other-fixtures→ chore(oie): ASVS 5.0 provisional B2 sources and gold #1117)stack/oie-p6-mapping-promotions→ chore(oie): NIST 800-53 and remaining B2 sources/gold #1118)stack/oie-p7-full-pipeline-docs→ feat(mappings): promote Related-orphan remaps and re-point the LLM fixture to AI-hub CREs (#1124) #1119)