Open Source Contribution Tracker
Evidence-backed public pull requests, reviews, CI signals, and maintainer follow-up.
View the Project on GitHub gnanirahulnutakki/oss-contribution-tracker
Open Source Contribution Tracker
A transparent, evidence-backed dashboard for my public open-source pull requests, reviews, and issue lanes. The goal is sustained usefulness to maintainers—not contribution-count inflation.
View the live dashboard or inspect every claim in this repository.
90-day scorecard
Program window: 2026-07-23 through 2026-10-20 (90 days). Outcome metrics include only new work from the program start; the authored-PR cap spans all active external work.
| Outcome | Current | Target | Progress |
|---|---|---|---|
| Merged external pull requests | 0 | 24 | 0% |
| Landing rate for decided PRs | — | 70% | No decided PRs yet |
| External pull requests reviewed | 27 | 12 | Target met |
| Simultaneous open authored PRs | 7 | ≤ 4 | Over cap |
Authored external PRs opened in this window: 5. Administrative gates excluded from landing-rate decisions: 1.
Fresh unsolicited review cadence: eligible under the automated guardrail (rolling 24-hour cap: 2; minimum spacing: 4 hours). Requested follow-ups remain evidence-driven and exempt.
Tracked portfolio
| Contribution | Role | Stage | Verification signals | Next action |
|---|---|---|---|---|
| prometheus/alertmanager#5402 victorops: use Splunk On-Call branding |
author | Awaiting CI approval | 2 checks passed · workflow approval needed · base advanced (update optional) | Wait for maintainer workflow approval; do not ping prematurely. |
| prometheus-community/postgres_exporter#1351 fix: preserve precision for large counters |
author | Changes requested | 1 active change request · 15 non-failing checks (13 passed, 2 skipped) · base advanced (update optional) | The CLI-only change is pushed in 9c8b1d9, its review thread is resolved, and all 15 checks are non-failing (13 passed, 2 skipped); wait for maintainer re-review. |
| PrefectHQ/fastmcp#4625 fix: use Python field name for structured task results |
author | Closed | 10 non-failing checks (9 passed, 1 skipped) · expected CI gate | Issue #4616 was closed by merged maintainer PR #4657; this administrative pull request remains closed. No follow-up is needed. |
| open-telemetry/otel-arrow#3572 fix(filter): reject unknown config fields |
author | PR open | 78 non-failing checks (73 passed, 5 skipped) · expected CI gate · base advanced (update optional) | Link and verify gnani.nutakki@gmail.com on GitHub, then complete EasyCLA; the PR is ready and twice approved with 73 passing and 5 skipped technical checks, but currently has no assignee. Do not post another routing comment before the legal gate is green. |
| envoyproxy/gateway#9574 docs: explain backend protocol selection |
author | PR open | 28 non-failing checks (14 passed, 11 skipped, 3 neutral) · base advanced (update optional) | Ready on unchanged head 12202063 after all 28 checks completed non-failing and Codex found no major issues. GitHub denied the fork author's Copilot review request for lack of repository permission; wait for human review or a concrete request without pinging. |
| oam-dev/cluster-gateway#171 fix(proxy): enforce managed credential precedence |
author | PR open | 8 checks passed | All eight checks pass and no review thread or maintainer reply is outstanding; wait without pinging. |
| NousResearch/hermes-agent#10824 fix(installer): reuse existing newer Python before downloading 3.11 |
author | PR open | 38 non-failing checks (27 passed, 10 skipped, 1 neutral) · base advanced (update optional) | Head ce1cd22 is rebased onto current main, bounds reuse to >=3.11,<3.14, preserves UV_PYTHON repinning, and passes 41 shell-installer plus 13 contributor-mapping tests. The review thread is resolved and all 27 reporting CI checks pass; wait for maintainer re-review without pinging. |
| NousResearch/hermes-agent#10828 fix(runtime): auto-upgrade GPT-5 named custom providers to codex_responses |
author | Closed | No checks reported | Closed after confirming model-only Responses inference conflicts with the current custom-provider compatibility policy; no follow-up is needed. |
| NousResearch/hermes-agent#10852 fix(gateway): keep auto vision preprocess concise |
author | PR open | 38 non-failing checks (26 passed, 11 skipped, 1 neutral) · base advanced (update optional) | Rebased head 07246b88 onto current main and preserved the explicit 500-token cap through custom-provider async retry/fallback paths. All 277 expanded local tests pass, and all 38 hosted checks are non-failing (26 passed, 11 skipped, 1 neutral), including amd64 and arm64 Docker builds; wait for maintainer re-review without posting a status comment. |
| prometheus-operator/prometheus-operator#8695 alertmanager: preserve top-level event_recorder configuration |
reviewer | Review submitted | 22 non-failing checks (21 passed, 1 skipped) | Watch the unresolved TLS compatibility finding and respond if the author follows up. |
| prometheus/client_golang#2062 api: return HTTP status on non-JSON API error bodies |
reviewer | Review submitted | 15 non-failing checks (14 passed, 1 skipped) | Watch the valid-JSON error-envelope finding and re-review if the author updates the pull request. |
| helm/helm#32339 fix(scripts): add cache-busting to Helm version checks |
reviewer | Review landed | 7 checks passed | Merged on exact reviewed head bc0114c at 2026-07-28T21:30:38Z after seven passing checks. A maintainer noted a possible dev-v3 backport but did not request action from this reviewer; do not reply or open follow-up work while over cap. |
| openai/openai-agents-python#3933 fix: enforce realtime text guardrails and synchronize streaming cancellation |
reviewer | Review complete | 10 non-failing checks (9 passed, 1 skipped) | Closed unmerged on 2026-08-02. The maintainer's request to stop posting general comments and identify only specific points is acknowledged; no reply or follow-up is warranted. |
| google/adk-python#6459 fix: route MCP calls to the .mtls.googleapis.com endpoint for Agent Identity |
reviewer | Review complete | 23 non-failing checks (19 passed, 4 skipped) · expected CI gate | Closed unmerged after a maintainer confirmed the endpoint rewrite already occurs in _get_connection_uri(). The explanation is specific and requires no reviewer reply or replacement pull request. |
| google/adk-python#6460 fix(litellm): strip embedded thought_signature from tool call id |
reviewer | Review update available | 6 non-failing checks (2 passed, 4 skipped) · expected CI gate · 1 failing · workflow approval needed | Head 7c6904ab still has the unresolved empty embedded-signature ID finding; the failing agent-triage run is an unrelated Gemini 503 and CLA remains the expected gate. Do not duplicate the inline finding; re-review only after a substantive author update. |
| pydantic/pydantic-ai#6706 Validate native tool IDs per capability layer |
reviewer | Review update available | 65 non-failing checks (48 passed, 16 skipped, 1 neutral) | Exact head a51fa5e7 still rejects conflicting native tools returned by a factory inside CombinedCapability even though the same tools override as separate run-level entries. Seven focused and 417 full agent tests pass, but the validation-boundary contract remains unresolved; GitHub returned 'User is blocked' when the proof-backed follow-up was submitted, so do not retry through another channel unless access changes or a maintainer contacts us. |
| microsoft/agent-framework#7292 Python: [Feature]: Support OpenAI instructions in Responses API |
reviewer | Review landed | 37 non-failing checks (23 passed, 14 skipped) | Merged on head 8a6a52c9 after two maintainer approvals and 37 non-failing checks (23 passed, 14 skipped); no follow-up required. |
| microsoft/autogen#7994 fix(autogen-ext): skip LangChain callback-manager (run_manager) when inferring tool args schema |
reviewer | Review submitted | 2 checks passed · workflow approval needed | Answered the author's concrete re-review question at issuecomment-5173162647: head 7b5e64de is unchanged from the independently tested and approved commit, so the existing approval still applies. Wait for maintainer workflow approval or a substantive new head. |
| crewAIInc/crewAI#6625 fix(reasoning): use flexible regex to detect READY state to prevent agent deadlocks |
reviewer | Review submitted | 1 check passed · workflow approval needed | Watch the semantic readiness-marker finding and re-review after the author updates head d2c0b593. |
| huggingface/smolagents#2565 fix: filter TOOL_CALL and TOOL_RESPONSE messages from managed-agent summary |
reviewer | Review submitted | workflow approval needed | Changes requested on exact head 288b92ef: ActionStep still serializes raw arguments and formats observations before the role filter, while observations_images become USER messages and bypass it; deterministic probes reproduced both raw-data access and an image marker leaking into the parent summary. Six focused tests and Ruff lint pass, but make quality fails formatting. Re-review only after redaction moves before rendering with argument-serialization and image regressions. |
| strands-agents/harness-sdk#3448 feat(graph): add Python concurrency limit |
reviewer | Review submitted | 33 non-failing checks (29 passed, 4 skipped) · 1 pending | The positive-integer validation fix is verified and approved on head 97c2b6e; wait for maintainer workflow authorization, review, or another head update. |
| openai/openai-python#3538 fix(streaming): return final response for incomplete and failed events |
reviewer | Review submitted | workflow approval needed | The incomplete/failed structured-output fix is verified on head e896c03a; wait for maintainer review or another head update. |
| langchain-ai/deepagents#5026 fix(sdk): make BackendProtocol.glob recursive for bare patterns |
reviewer | Review submitted | 62 non-failing checks (44 passed, 18 skipped) · 3 failing | The original expansion DoS is fixed on b965c14b, but {{a,b}}.py disagrees between sandbox and shared backends and the Windows test expects backslashes despite production POSIX normalization; that assertion is the sole real hosted failure. GitHub denied a formal changes-requested review without explicit repository access, so a proof-backed COMMENT review was submitted. Re-review after both regressions change. |
| sgl-project/sglang#32344 [Bugfix] Emit cached token metric before first cache hit |
reviewer | Review submitted | 85 non-failing checks (6 passed, 79 skipped) · expected CI gate | Watch the PromQL label-set mismatch finding on head 94b638a4; re-review after the metric schema or query and regression coverage are corrected. |
| browser-use/browser-use#5300 Call the model with minimal state after capture timeout |
reviewer | Review landed | 96 non-failing checks (94 passed, 2 skipped) | Watch for author or maintainer follow-up and re-review only if the head changes or a concrete timeout-safety question is raised. |
| modelcontextprotocol/typescript-sdk#2544 Fix tests timing out during interactive OAuth flow (#2510) |
reviewer | Review update available | 14 checks passed | Watch the unresolved REDIRECT lifecycle finding on head a740d28c; re-review after the triggering request stays pending through finishAuth and retry with regression coverage. |
| confident-ai/deepeval#2950 fix(openai): apply prompt-caching discount to GPTModel cost calculation |
reviewer | Review submitted | 12 checks passed · expected CI gate · 2 failing | Re-tested exact head a2758d42 and replied in the existing unresolved thread. Standard cached rates are fixed, but fully cached 300K-token GPT-5.4 input returns $0.075 instead of the documented $0.15 because the >272K long-context multiplier is absent; the same reproduces on current main and 52 affected tests pass without a boundary regression. Wait for a substantive fix; lint is current-main drift and Vercel needs team authorization. |
| traceloop/openllmetry#4375 fix(cohere): avoid double-ending async error spans |
reviewer | Review submitted | 3 checks passed · workflow approval needed | Watch the duplicate exception-event finding on head 9c303fd1; re-review after one layer owns exception recording and the regression asserts exact cardinality. |
| Arize-ai/openinference#3429 fix(anthropic): preserve APIPromise helpers |
reviewer | Review submitted | 21 non-failing checks (14 passed, 7 skipped) | The consumed asResponse body finding is verified fixed and approved on head d4827bfb; all 16 package tests, type-check, build, lint, and format passed, and all 21 GitHub checks completed without failure. Wait for maintainer action or a changed head. |
| AgentOps-AI/agentops#1428 fix(langchain): capture structured output tool calls |
reviewer | Review submitted | 1 check passed · workflow approval needed | Structured-output capture is reproduced and approved on head 85f6896; wait for maintainer review, CI authorization, or a new head. |
| agno-agi/agno#9174 refactor: cache the pydantic version across tool wrappers |
reviewer | Review complete | 7 non-failing checks (5 passed, 2 skipped) · 7 failing | Closed as superseded after #9210 landed the same cache on main. The author acknowledged the independent validation and asked no question; no reply or re-review is warranted. |
| omnigent-ai/omnigent#3304 fix(accounts): enforce the last-admin invariant atomically on delete |
reviewer | Review landed | 68 non-failing checks (60 passed, 8 skipped) | Merged on exact reviewed head bb2ca4c at 2026-07-29T00:07:09Z after maintainer approval and 68 non-failing checks (60 passed, 8 skipped). The approval thanked the contributor but asked no question; no reply is warranted. |
| opensquilla/opensquilla#815 fix(engine): evict cache-break monitor state for terminal sessions |
reviewer | Review submitted | 22 non-failing checks (18 passed, 4 skipped) | Re-approved exact head ef0a52b3 after two main merges: the conflict resolution correctly moved cache-break eviction into the renamed evict_session_runtime_state hook before the generation-specific session_id guard while preserving finish and explicit history-deletion coverage. All 19 focused tests, Ruff on the four changed files, and mypy on both changed source files pass; GitHub reports 18 successful and 4 skipped checks. Wait for maintainer action or another substantive head. |
| agentscope-ai/agentscope#1874 fix(embedding): omit default OpenAI dimensions |
reviewer | Review submitted | 6 checks passed | Two P2 findings are open on head ce2832e: preserve explicit pass_dimensions=False when Parameters uses defaults, and retain genuine multi-type anyOf schemas. Re-review only after an author update or concrete reply. |
| tma1-ai/tma1#73 chore(deps): Bump sharp and astro in /site |
reviewer | Review submitted | 1 check passed | Exact head afa47ab builds on Node 22.12.0 and preserves the three localized page contracts, but the lockfile still installs sharp 0.34.5 and leaves GHSA-f88m-g3jw-g9cj present. One file-level finding is open; wait for a head update or concrete reply without duplicating it. |
| kserve/kserve#5917 fix: select compatible LocalModelNodeGroup instead of blindly picking [0] |
reviewer | Review submitted | 5 checks passed · workflow approval needed | Re-reviewed exact head 4c1d4529 after the RBAC and generated-manifest update: controller-gen, Helm, Kustomize, all 12 quick-install outputs, 21 validators, race tests, compile/vet, chart lint, and diff checks pass. GitHub denied a formal approval because the account lacks explicit repository access, so a proof-backed COMMENT review was submitted; wait for maintainers to authorize the remaining action_required workflows. |
| oam-dev/cluster-gateway#170 Security: managed credential must override inbound Authorization on proxy requests |
author | Issue open | unassigned | The linked implementation PR is green; wait for maintainer review without a status ping. |
CaviraOSS/OpenMemory#186delete_all removes DB rows but not the in-process hsg query cache — deleted memories remain retrievable for the process lifetime |
author | Assigned elsewhere | assigned to @nullure | Resolved upstream by merged competing PR #190, which added cache clearing directly. The issue remains open and assigned elsewhere, but no implementation or reply is owed from this contributor. |
| plastic-labs/honcho#790 Deleted workspace's search results keep being served from the redis cache until TTL; recreating the same workspace id resurrects them |
author | Issue open | unassigned | No maintainer reply or assignment has landed; preserve the report and do not ping. |
| shaneholloman/mcp-knowledge-graph#22 Memories silently written to a global shared store when cwd lacks project markers — cross-project contamination by default; --memory-path ignored |
author | Issue open | unassigned | No maintainer reply or assignment has landed; preserve the report and do not ping. |
| akuity/kargo#6685 Webhooks server (kubernetes-webhooks-server) is missing probe support that all other servers have |
assignee | Assigned | assigned to @gnanirahulnutakki | Assigned implementation is valid, but hold publication while the global authored-PR cap is exceeded; recheck for competing work before resuming. |
| kagent-dev/kagent#2303 A2A: report turn-total token usage on the terminal status update |
participant | Issue open | unassigned | Issue author @QuentinBisson opened the exact implementation as draft PR #2332 on 2026-07-25; do not implement, claim, review, or comment. Remove this lane from the candidate queue unless a maintainer makes a concrete request. |
| PrefectHQ/fastmcp#2879 Background job does not produce valid CreateTaskResult |
participant | Issue open | unassigned | No author or maintainer reply followed the prior technical comment; retain as a watch-only legacy lane. |
| argoproj/argo-cd#21059 CLI: add possibility to diff the desired state instead of the live state |
participant | Issue open | unassigned | Maintainer PR #27469 implements this lane. Its current head has older approvals and green completed checks but is now dirty and review-required; preserve the prior report, do not compete, and do not comment unless a maintainer asks a concrete question. |
| argoproj/argo-cd#12273 Rework Credentials template url layout in Settings: Repositories |
participant | Issue open | unassigned | Our implementation PR #27415 was self-closed and later competing attempts also closed. Keep this as watch-only legacy context; do not revive it without fresh maintainer intent and current reproduction. |
| NousResearch/hermes-agent#10695 Python dependency CVEs: aiohttp, cryptography, curl-cffi need minimum version bumps |
participant | Issue closed | unassigned | Closed after alternative PR #26830 merged. The maintainer explicitly credited this contributor's audit; no reply or replacement work is needed. |
| meshery/meshery#18739 [UI] Replace direct MUI imports in dashboard surfaces |
assignee | Issue closed | assigned to @leecalcote, @gnanirahulnutakki | Closed after maintainer PR #19306 merged and credited this contributor's initial PR #18770. No reply or competing implementation is needed. |
| akuity/kargo#6021 feat: add gitea label support |
participant | Issue closed | assigned to @fuskovic | Closed after maintainer PR #6262 merged; this contributor's earlier PR #6101 was self-closed. No follow-up is needed. |
| kitops-ml/kitops#1138 Set org.opencontainers.image.created annotation on ModelKit manifests |
participant | Issue closed | unassigned | Closed after competing PR #1139 merged on 2026-05-11; this contributor's PR #1164 must remain closed and no reply is needed. |
| MemoriLabs/Memori#590 Self-hosted (BYODB) usage: capture_agent_turn() POSTs to the cloud API, and Memori(conn=...) never provisions the schema — both undocumented |
author | Issue closed | unassigned | Closed after the current-version clarification. The later DevRel Discord invitation contains no technical question, so no public reply or issue revival is warranted. |
| basicmachines-co/basic-memory#944 CLI write-note (and other tool subcommands) crash with 'function' object has no attribute 'fn' — no resolvable fastmcp version satisfies the 0.16.2 CLI bridge |
author | Issue closed | unassigned | Closed after the maintainer confirmed the behavior is fixed in current releases. Maintainer-authored issue #1181 later cross-referenced #944 only as an older resolved FastMCP 2.x/3.x pin issue that does not overlap the FastMCP 4 beta port; the non-actionable reference is acknowledged and requires no reply or revival. |
| kagent-dev/kagent#1037 feat: add support for Docker images and kmcp projects for local dev |
participant | Issue closed | unassigned | Closed by the stale bot on 2026-06-23 without a maintainer response. Do not revive the old claim without current reproduction and renewed maintainer intent. |
| letta-ai/letta#3364 BYOK OpenAI-compatible provider: handle is generated under openai-proxy/... but validated against the provider name → 400, or unhandled AssertionError → opaque 500 |
author | Issue closed | unassigned | Automatically closed and locked for missing the repository's required human-verification and AI-policy template. Do not reopen or repost without a fresh reproduction and full policy compliance. |
Last public state change recorded: 2026-08-04T15:11:24+00:00.
The scheduled job still checks daily. It commits only when these public facts change.
Operating principles
- Quality before volume: every implementation must be scoped, tested, and reviewable.
- Maintainer attention is scarce: no duplicate pull requests, assignment nudges, or premature follow-ups.
- Work in progress stays bounded to four authored pull requests, with only one active implementation per project.
- External reviews count when they contain a reproduced, actionable finding.
- Fresh unsolicited reviews are limited to one per automated cycle and two per rolling 24 hours, with at least four hours between them.
- Automated checks commit only when real public state changes; daily no-op refreshes do not manufacture profile activity.
The complete daily loop and intake rules are in docs/CONTRIBUTION_PLAYBOOK.md.
How it works
data/contributions.json is the small, human-reviewed portfolio. A scheduled
GitHub Actions workflow reads public GitHub API data, updates
data/snapshot.json, and regenerates only the marked dashboard section above.
It uses the repository’s short-lived GITHUB_TOKEN; no personal access token
or third-party credential is required.
The refresh records exact heads, draft state, base-branch drift, and whether that drift requires action; detects active change requests on authored pull requests; and surfaces only direct human mentions, replies to the contributor’s own review threads, or trusted linked-issue responses after the contribution began. Base drift is bucketed instead of counted commit by commit, and optional drift is kept distinct from repository-enforced branch updates to avoid needless CI churn. Active legacy pull requests and issue discussions remain visible even when they predate the scorecard window. All open external authored pull requests count against the simultaneous-work cap; only outcome metrics stay window-bounded. The tracker also derives the next fresh-review eligibility boundary from first-submission timestamps while leaving explicitly requested follow-ups exempt. Bot-only pull-request timestamp churn is stabilized so it cannot create synthetic tracker commits. Human-reviewed, non-actionable response signals can be acknowledged per contribution without suppressing later activity. The tracker never posts upstream comments.
For snapshot compatibility, checks.passing keeps its original meaning:
successful, skipped, and neutral checks. Consumers that need exact evidence
should use checks.successful for successful checks and checks.non_failing
for the explicit aggregate.
Reuse this tracker
This repository is designed to work as a GitHub template. Configuration lives
in data/contributions.json; the update engine and workflows do not depend on
the original repository owner.
Follow the setup guide to configure a reporting window, add reviewed contributions, enable the scheduled refresh, and verify that unchanged public state produces no synthetic commit.
Run it locally:
GITHUB_TOKEN="$(gh auth token)" python3 scripts/update_tracker.py
python3 -m unittest discover -s tests -v
Add a contribution
Add a public pull request or review entry to data/contributions.json, then
run the updater and tests. The schema is deliberately plain JSON so every
tracked claim remains easy to inspect and change.