chore: migrate to HodeTech org + ignore new nightly CVE#61
Open
cemililik wants to merge 3 commits into
Open
Conversation
The project moved to the HodeTech GitHub organization (github.com/HodeTech/ForgeLM). Update every reference project-wide: - Repo URLs: github.com/cemililik/ForgeLM -> github.com/HodeTech/ForgeLM (README badges, pyproject [project.urls], docs, user manuals EN+TR, notebooks, site/, CHANGELOG, CONTRIBUTING, issue-template config, model_card.py, standards, roadmap). - GHCR image: ghcr.io/cemililik/forgelm -> ghcr.io/hodetech/forgelm (lowercase namespace, per GHCR owner naming). - Contact email (CODE_OF_CONDUCT.md): ai-team@cemililik.com -> forgelm@hodetech.io. - SonarCloud (sonar-project.properties): organization=hodetech, projectKey=HodeTech_ForgeLM. 271 references across 99 files, 1:1 replacements (no other edits). Local git remote origin re-pointed to the HodeTech URL. Doc/site guards re-run green: anchor resolution, bilingual parity, user-manual self-contained, CLI help consistency, site-version, site-claims. model_card tests pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…5-24) The nightly on main is green through the new --ignores mechanism (the run log shows CVE-2026-1839 correctly suppressed), but a new no-fix advisory dropped: transformers PYSEC-2025-217 (CVE-2025-14929), X-CLIP checkpoint-conversion deserialization RCE. CVSS is AV:L/UI:R (local + user must open a malicious checkpoint), affected through 5.0.0rc0 with no fix in the pinned 4.x line. Same threat-model carve-out as the existing transformers/torch deserialization ignores: operators are documented as required to load only trusted checkpoints, and ForgeLM does not invoke X-CLIP checkpoint conversion. With the Phase-refactored ignore file this is a one-line data change, no workflow edit — exactly the maintainability win the PR #59 review asked for. Refs: #58 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
cemililik
changed the title
Collaborator
Author
|
@coderabbitai review all |
|
✅ Actions performedFull review triggered. |
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (4)
📝 WalkthroughWalkthroughThis PR replaces repository/registry/organization references from cemililik/ForgeLM to HodeTech/ForgeLM across docs, site, notebooks, CI, project metadata, and adds a pip-audit ignore entry plus deployment-note and exit-code clarifications. ChangesRepository & CI migration
Documentation & manuals
Code, notebooks, model card
Website & translations
Feature notes & contracts
Security vulnerability management
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Warning There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure. 🔧 ast-grep (0.42.3)site/js/translations.jsComment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Nitpick comments (1)
docs/usermanuals/en/deployment/chat.md (1)
110-110: ⚡ Quick winMisleading equivalence between chat-compare and benchmark-only.
The note states "Today the same comparison runs through
forgelm --benchmark-only", but the--benchmark-onlymode runs lm-evaluation-harness benchmark tasks (forgelm/cli/_no_train_modes.py:1-140), which is fundamentally different from the side-by-side chat prompt comparison shown in lines 113-119. Consider revising to acknowledge that no direct equivalent exists today, or clarify that operators must manually run benchmarks separately on each model and compare the judge results themselves (which is a different workflow from the planned dedicated UX).🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@docs/usermanuals/en/deployment/chat.md` at line 110, The note incorrectly implies feature parity between the planned chat-compare subcommand and the existing --benchmark-only mode; update the text to state that chat-compare provides an interactive side-by-side chat prompt comparison while --benchmark-only (see forgelm/cli/_no_train_modes.py) runs lm-evaluation-harness benchmarks and requires manually running benchmarks per checkpoint and aggregating judge results, so operators must currently run and compare results themselves rather than getting the planned dedicated UX.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/usermanuals/en/deployment/chat.md`:
- Around line 66-68: Resolve two issues: pick one consistent flag form (e.g.,
use "--safety" or "--safety on" everywhere) and update the paragraph to use that
chosen form; and clarify the current v0.5.5 behavior by either (A) describing
the exact binding workflow (pointing readers to the chat subcommand parser in
forgelm/cli/_parser.py and specifying which config file/key to set and how the
runner must load it) or (B) explicitly state that on v0.5.5 the CLI does not
support per-turn safety screening and that the YAML safety: enabled: true entry
is only read by certain runners (name them) — do not claim a binding mechanism
exists unless you spell out the exact steps referencing forgelm/cli/_parser.py.
In `@site/js/translations.js`:
- Line 4345: The privacy copy currently hardcodes "pip install forgelm==0.5.5"
in the "privacy.body" locale string (and the same hardcoded version appears in
the sibling privacy locale entries noted in the comment); change this to avoid a
stale pinned version by either removing the "==0.5.5" (use "pip install
forgelm") or replace it with a template/variable (e.g., "{FORGELM_PIP}" or
"{siteVersion}") that the site build injects so the install line is maintained
site-wide and updated from a single source; locate and update the string key
"privacy.body" (and the other privacy locale keys mentioned) to use the
non-hardcoded form.
In `@tools/pip_audit_ignores.yaml`:
- Around line 51-69: Update the PYSEC-2025-217 ignore entry (id: PYSEC-2025-217,
package: transformers) to match advisory wording: state CVE-2025-14929 with CVSS
vector including AV:L and UI:R, remove the unverified phrase "affected through
5.0.0rc0" and instead say "no fixed version announced for the 4.x line" (or
mirror the advisory language about lack of a fix), and adjust the
reevaluate_after criteria accordingly; additionally perform and document a
codebase check to confirm the threat-model claim by searching for usages of
X-CLIP checkpoint conversion paths (search for identifiers/strings such as
"X-CLIP", "checkpoint conversion", "pt2 loader", and any direct "torch.load"
checkpoint-loading sites) and record the result in the entry.
---
Nitpick comments:
In `@docs/usermanuals/en/deployment/chat.md`:
- Line 110: The note incorrectly implies feature parity between the planned
chat-compare subcommand and the existing --benchmark-only mode; update the text
to state that chat-compare provides an interactive side-by-side chat prompt
comparison while --benchmark-only (see forgelm/cli/_no_train_modes.py) runs
lm-evaluation-harness benchmarks and requires manually running benchmarks per
checkpoint and aggregating judge results, so operators must currently run and
compare results themselves rather than getting the planned dedicated UX.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 3115152d-2066-4049-a433-0504b4e6c23f
📒 Files selected for processing (100)
.agents/skills/cut-release/SKILL.md.agents/skills/sync-bilingual-docs/SKILL.md.claude/skills/cut-release/SKILL.md.claude/skills/sync-bilingual-docs/SKILL.md.github/ISSUE_TEMPLATE/config.yml.github/workflows/nightly.ymlCHANGELOG.mdCODE_OF_CONDUCT.mdCONTRIBUTING.mdREADME.mddocs/guides/quickstart-tr.mddocs/guides/quickstart.mddocs/guides/troubleshooting-tr.mddocs/guides/troubleshooting.mddocs/reference/usage-tr.mddocs/reference/usage.mddocs/roadmap/releases.mddocs/roadmap/risks-and-decisions.mddocs/standards/code-review.mddocs/standards/documentation.mddocs/usermanuals/en/compliance/annex-iv.mddocs/usermanuals/en/compliance/audit-log.mddocs/usermanuals/en/compliance/gdpr-erasure.mddocs/usermanuals/en/compliance/gdpr.mddocs/usermanuals/en/compliance/human-approval-gate.mddocs/usermanuals/en/compliance/overview.mddocs/usermanuals/en/compliance/verify-audit.mddocs/usermanuals/en/data/deduplication.mddocs/usermanuals/en/data/pii-masking.mddocs/usermanuals/en/deployment/chat.mddocs/usermanuals/en/deployment/model-merging.mddocs/usermanuals/en/deployment/verify-gguf.mddocs/usermanuals/en/evaluation/trend-tracking.mddocs/usermanuals/en/getting-started/installation.mddocs/usermanuals/en/operations/docker.mddocs/usermanuals/en/operations/experiment-tracking.mddocs/usermanuals/en/operations/gpu-cost.mddocs/usermanuals/en/operations/iso-soc2-deployer.mddocs/usermanuals/en/operations/supply-chain.mddocs/usermanuals/en/operations/troubleshooting.mddocs/usermanuals/en/operations/webhooks.mddocs/usermanuals/en/reference/configuration.mddocs/usermanuals/en/reference/exit-codes.mddocs/usermanuals/en/reference/json-output.mddocs/usermanuals/en/reference/library-api.mddocs/usermanuals/en/reference/performance.mddocs/usermanuals/en/reference/yaml-templates.mddocs/usermanuals/tr/compliance/annex-iv.mddocs/usermanuals/tr/compliance/audit-log.mddocs/usermanuals/tr/compliance/gdpr-erasure.mddocs/usermanuals/tr/compliance/gdpr.mddocs/usermanuals/tr/compliance/human-approval-gate.mddocs/usermanuals/tr/compliance/overview.mddocs/usermanuals/tr/compliance/verify-audit.mddocs/usermanuals/tr/data/deduplication.mddocs/usermanuals/tr/data/pii-masking.mddocs/usermanuals/tr/deployment/chat.mddocs/usermanuals/tr/deployment/model-merging.mddocs/usermanuals/tr/deployment/verify-gguf.mddocs/usermanuals/tr/evaluation/trend-tracking.mddocs/usermanuals/tr/getting-started/installation.mddocs/usermanuals/tr/operations/docker.mddocs/usermanuals/tr/operations/experiment-tracking.mddocs/usermanuals/tr/operations/gpu-cost.mddocs/usermanuals/tr/operations/iso-soc2-deployer.mddocs/usermanuals/tr/operations/supply-chain.mddocs/usermanuals/tr/operations/troubleshooting.mddocs/usermanuals/tr/operations/webhooks.mddocs/usermanuals/tr/reference/configuration.mddocs/usermanuals/tr/reference/exit-codes.mddocs/usermanuals/tr/reference/json-output.mddocs/usermanuals/tr/reference/library-api.mddocs/usermanuals/tr/reference/performance.mddocs/usermanuals/tr/reference/yaml-templates.mdforgelm/model_card.pynotebooks/data_curation.ipynbnotebooks/dpo_alignment.ipynbnotebooks/galore_memory_optimization.ipynbnotebooks/grpo_reasoning.ipynbnotebooks/ingestion_playground.ipynbnotebooks/kto_binary_feedback.ipynbnotebooks/multi_dataset.ipynbnotebooks/post_training_workflow.ipynbnotebooks/quickstart_sft.ipynbnotebooks/safety_evaluation.ipynbnotebooks/synthetic_data_training.ipynbpyproject.tomlsite/README.mdsite/compliance.htmlsite/contact.htmlsite/features.htmlsite/guide.htmlsite/index.htmlsite/js/guide.jssite/js/translations.jssite/privacy.htmlsite/quickstart.htmlsite/terms.htmlsonar-project.propertiestools/pip_audit_ignores.yaml
…acy pin, PYSEC-2025-217) All four findings verified against current code; all still valid. chat.md (EN+TR) — safety routing (L66-68): Verified forgelm/cli/_parser.py: `forgelm chat` has no --safety flag and loads no config; run_chat/ChatSession take no safety arg. The old copy claimed a YAML `safety:` binding into the chat session — false, and the key path was wrong. Rewritten to state the current release does NOT screen per-turn in the REPL, and that `evaluation.safety.enabled: true` is consumed only by the training auto-revert gate (forgelm/trainer.py) and `forgelm safety-eval`. Collapsed the inconsistent `--safety` / `--safety on` forms to a single `--safety`. chat.md (EN+TR) — chat-compare note (L110): Verified forgelm/cli/_no_train_modes.py: --benchmark-only runs lm-evaluation-harness on a SINGLE checkpoint, no side-by-side, no judge. Removed the implied parity; now says operators must run --benchmark-only per checkpoint, run the judge separately, and aggregate themselves until the planned chat-compare UX ships. site/js/translations.js — privacy.body (6 locales): Dropped the stale hard-coded `pip install forgelm==0.5.5` pin (real release is 0.7.0; this string is outside update_site_version.py's scope so it drifted). Now `pip install forgelm` — no version to drift, consistent with the site's no-build-step constraint. tools/pip_audit_ignores.yaml — PYSEC-2025-217: Reworded to advisory language: full CVSS vector (AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H), "no fixed version for the 4.x line" instead of "affected through 5.0.0rc0", reevaluate_after no longer assumes a fixed 5.x exists. Documented the codebase check (git grep over forgelm/): no X-CLIP usage, no checkpoint-conversion path besides HF->GGUF export, zero direct torch.load sites — all loading via from_pretrained, so the vulnerable path is unreachable. Validated: bilingual parity, anchor resolution, user-manual self-contained, site-version, site-claims, site chrome parity, CLI help consistency, ignore-file schema; 23 check_pip_audit tests pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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
1 participant
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
Two related housekeeping changes that both target
mainviadevelopment:github.com/HodeTech/ForgeLM). Updates every reference project-wide (271 refs / 99 files, 1:1 string replacements).transformers PYSEC-2025-217to the project ignore list so the daily nightly goes green again.1. Org migration
github.com/cemililik/ForgeLM→github.com/HodeTech/ForgeLM(README badges,pyproject.toml[project.urls], docs, user manuals EN+TR, notebooks,site/, CHANGELOG, CONTRIBUTING, issue-template,forgelm/model_card.py, standards, roadmap)ghcr.io/cemililik/forgelm→ghcr.io/hodetech/forgelm(lowercase namespace)ai-team@cemililik.com→forgelm@hodetech.io(CODE_OF_CONDUCT.md)organization=hodetech,projectKey=HodeTech_ForgeLM(sonar-project.properties)git grep cemililik→ 0 remaining. Guards green: anchor resolution, bilingual parity, user-manual self-contained, CLI help consistency, site-version, site-claims.ruffclean;model_cardtests pass.2. Nightly CVE ignore (transformers PYSEC-2025-217)
The new
--ignores tools/pip_audit_ignores.yamlmechanism works onmain(the run log showsCVE-2026-1839correctly suppressed), but a fresh no-fix advisory dropped:AV:L/UI:R(local + user must open a malicious checkpoint), affected through5.0.0rc0, no fix in the pinned 4.x line.The ignore-file loader self-validates (
tools/check_pip_audit.py --ignores); 23 tests pass.Repo-side follow-ups (not in this diff)
hodetechorg with keyHodeTech_ForgeLM; re-addSONAR_TOKENsecret if not inherited (repo currently lists no Actions secrets).Closes #58
🤖 Generated with Claude Code
Summary by CodeRabbit
Chores
Documentation
Security