Claude Project Tracker

Accepts GitHub webhook events (configure on the repo with a shared secret). When a pull_request event with action=closed and merged=true comes in, parse the PR body for `Closes #<item_id>` or `Fixes #<item_id>` patterns and call the existing complete_item path for each. Add: HMAC-SHA256 signature verification using `GITHUB_WEBHOOK_SECRET` env var. Route lives in main.go alongside the other public POST endpoints. Probably one new file github_webhook.go.

claude-project-tracker · claude-project-tracker · by saravanan@scrumclaw.ai

BRD v3.0 REQ-13.11 is Pending Deploy. Runbook with ready-to-paste gcloud commands exists; needs the Cloud Scheduler job created and 7 days of error-free runs to close the acceptance.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

RoadIntelix · roadintelix · by Prajith

RoadIntelix · roadintelix · by Prajith

Was deferred; now started. Same Expo codebase as iOS. DONE in code/config: eas.json Android build profiles (APK dev/preview, AAB production) + submit track:internal; app.json android block already complete (package com.hopewellpartners.adaptivesat, adaptive icon, App Links intent filters) — no change needed (appVersionSource:remote manages versionCode). Apple Sign-In correctly iOS-gated; Google sign-in works on Android. PENDING (owner): Google Play Developer account ($25); `eas build --platform android --profile preview` then production AAB; Android OAuth client (SHA-1 → add to GOOGLE_OAUTH_CLIENT_IDS); App Links assetlinks.json (SHA-256) on adaptivesat.ai; Play Console listing + IARC content rating (teen audience) + Data Safety + privacy URL; service account + eas submit. Follow-ons: Google Play Billing IAP (source='google' already supported), FCM push (deferred). Full steps: docs/ANDROID_PLAY_STORE_RUNBOOK.md.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

Apple rejected build 6 (reviewed on iPad Air M3, iPadOS 26.5.2). Four items. CODE FIXES DONE (mobile, typecheck clean, not yet committed/built): (1) 3.1.2(c) MobilePaywall now has tappable Terms of Use (EULA) + Privacy Policy links and shows title/length/price; (2) 2.1(b) SettingsScreen has a visible 'Upgrade to Premium' row opening the paywall so reviewers can locate the IAP; (3) 2.1(a) ExamStartScreen adds a 30s timeout + clear retry error so 'Start Exam' can't silently hang. ASC METADATA DONE: EULA link added to App Description (saved), Privacy Policy URL confirmed (adaptivesat.ai/privacy). PENDING (owner): check/top-up OpenAI credits (prime suspect for the iPad 'Start Exam did nothing' hang) and retest on iPad; commit+push mobile changes via ship.sh; new EAS production build (build 7); fresh screen recording (Start Exam works + Settings→Upgrade to Premium paywall + sandbox purchase); paste reply (asc_submission/apple_reply_draft.md) in Resolution Center; resubmit app+subscription+group together. Refs: MobilePaywall.tsx, SettingsScreen.tsx, ExamStartScreen.tsx.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

ProcessTopupWebhook uses `if topup.Amount != amount` — direct float64 equality. A tiny rounding drift (e.g., 500.00 stored vs 499.99999998 derived from Razorpay's paise/100) will falsely reject the webhook as 'amount mismatch', leaving the user paying but not credited. Change to: if math.Abs(topup.Amount - amount) > 0.01 { ... } Or better: store amount as int64 paise everywhere and compare integers.

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Migration 0027 only inserts positive ADJUSTMENT entries when `wallets.balance > 0 AND gap.amount >= 0.01`. Users whose legacy wallets.balance is 0 (never populated) but who accumulated CHARGE entries end up with negative ledger balances and no automatic reconciliation. Concrete case: saravanan.hp@gmail.com currently shows balance = -11152.52. Ledger has CHARGE entries but no matching TOPUP history was migrated in. Options: (a) Backfill missing legacy topups from external Razorpay reports before running the reconcile. (b) Add a companion migration that reconciles from a manually-curated CSV of legacy balances. (c) Add an admin credit tool + audit trail so support can fix these case-by-case. Track down the actual topup history for affected users (query Razorpay payments API filtered by our account) and decide the right long-term reconciliation approach.

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

After BRD + design review are locked, produce the formal technical design doc covering: - Sequence diagram (customer → Razorpay → backend → OCPP → charger) - Data model changes (schema migration text) - API contracts (webhook handler, admin QR provisioning, guest-mode read endpoints) - Idempotency + concurrency model - Failure/retry semantics (webhook re-fires, gun offline, RemoteStart rejected, refund flow) - SMS content + provider integration details - Observability (metrics + logs to add) - Test plan (unit + integration + one live pilot station) - Rollout / rollback plan - Estimated effort per phase Commit finalized TDD to docs/TDD-walkin-qr.md in the backend repo. Only after this ships do we start implementation items.</body>

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

DLT SMS registration is NOT done. 2–4 weeks external timeline. Track in parallel to Phase 1 build. Templates needed (send via Fast2SMS post-approval): - **wk_session_start** — "Vajra Volt: Charging started at {station}·Gun {gun}. ₹{reserved} reserved. Support: 88831-61155." - **wk_session_low_balance** — "Vajra Volt: Wallet running low (₹{remaining}). Scan QR to add money and keep charging." - **wk_session_end** — "Vajra Volt: Session complete. Used {kwh} kWh, cost ₹{cost}. Balance ₹{remaining}. Reply CLAIM to link this account." - **wk_payment_gun_busy** — "Vajra Volt: Payment ₹{amount} received but Gun {gun} is unavailable. Amount saved to your wallet. Support: 88831-61155." - **wk_refund_initiated** — "Vajra Volt: Refund of ₹{amount} initiated to your UPI. Reflects in 1–3 business days." All 4 languages (Q11 answer). English variants can be filed immediately; regional translations need vendor. Phase 1 launches WITHOUT SMS (see decision item). SMS goes live in Phase 1.5 once DLT approved.</body>

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

DLT registration is 2–4 weeks. Rather than delay walk-in launch, Phase 1 will ship with LED-based confirmation only: - Customer knows charging started when the green LED on the gun activates (Q5 confirmed). - No SMS on session start, low balance, or session end for Phase 1 walk-ins. - App users still get all their existing notifications (unaffected). - After DLT approval + template registration, Phase 1.5 patches SMS in as a config flip — no schema change needed. Trade-off: walk-in session-end UX is worse. If a customer left their car and walked away, they won't know when charging finished. Mitigation: - Set the 2-hour hard cap (Q6) so guns don't stay locked forever. - Print the phone number on the QR sticker so customers can call to check status. - Auto-stop threshold of ₹50 remaining (Q7) reduces the "session died silently" scenario for underpayers. If DLT lands earlier than expected we'll enable SMS immediately.</body>

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Pre-launch gate for Phase 1. Test matrix:\n\n- App: BHIM, Google Pay, PhonePe, Paytm, WhatsApp Pay (5 apps)\n- Amount: ₹100, ₹500, ₹1000 (3 amounts)\n- Scenario: happy path, gun offline (simulate), gun already charging, RemoteStart rejected (simulate)\n\nTotal ~30 test runs at pilot station. Log any UPI-app-specific quirks (BHIM in particular has stricter parsing).\n\nWhoever pilots — use throwaway UPI-linked phone number so we can also validate the walk-in-user creation path (not an existing app user's phone).\n\nBlocked on Phase 1 implementation complete + QR sticker printed.</body>

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Physical rollout for Phase 1 pilot.\n\n- Confirm which station is the pilot (default VAJRA0001)\n- Send print design (#56) to signage vendor (#Q17 answer: existing vendor)\n- Receive 2 laminated A5 stickers (Gun 1 + Gun 2)\n- Physically apply to gun/panel\n- Coordinate with ops staff to be on-site during pilot testing window\n- Document station-owner communications (landlord awareness of pilot)\n\nBlocked on #56 (design finalized) + Razorpay QR issued (#52 + #61).</body>

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Owner noticed the full exam has no back/forth between answered questions. Root cause: exam engine is forward-only by design. Backend get_next_exam_question serves the next UNSEEN question and immediately consumes it (status→ANSWERED); /answer records a single Attempt; no endpoint to re-fetch a prior question, change an answer, or list the module; ExamQuestionStatusEnum has only UNSEEN/ANSWERED/SKIPPED (no FLAGGED/REVIEW). Same limitation on WEB (frontend ExamSession.tsx) since it's the same backend — whole-product gap, not mobile-only. Real Digital SAT (Bluebook) allows free within-module navigation, changing answers, and a question navigator with mark-for-review until the module timer ends. SCOPE (multi-day, cross-cutting): backend — serve whole module, mutable per-question answers until module submit/timer, add flag/review state, add submit-module step (stop consume-on-serve); web+mobile — question navigator UI (grid + flags), Back/Next, module answer state, submit at module end. NOT an App Store blocker — sequence AFTER build 7 clears review. Refs: backend/app/services/exam_service.py (get_next_exam_question, submit_exam_answer), backend/app/models/exam_module_question.py, mobile/src/screens/ExamSessionScreen.tsx, frontend/src/app/pages/ExamSession.tsx.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

VajraCMS went live at https://cms.vajravolt.com on 2026-09-07. The seed account created by VajraBackendGo/migrations/0033_cms_user.up.sql is admin@vajravolt.in / admin123 — the plaintext password is written in the migration, so anyone with repo access has working credentials to the production CMS. Saravanan chose to defer this on the day. Not blocking anything, but it should not sit for long. If there is no change-password UI yet, on the VM: UPDATE "CMSUser" SET password_hash = crypt('NEW', gen_salt('bf',10)) WHERE email='admin@vajravolt.in'; (pgcrypto bcrypt matches what the CMS auth controller compares against) Verify the old one stops working: curl -s -o /dev/null -w '%{http_code}\n' -X POST https://cms-api.vajravolt.com/api/v1/auth/login \ -H 'Content-Type: application/json' -d '{"email":"admin@vajravolt.in","password":"admin123"}' Expect 401.</body>

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Found 2026-09-26. tracker_get_log with no project_id returns newest = #20 (adaptive-sat, 2026-07-23). Updates #21–#27 exist and show correctly per project (vajra-mobile-app #21–#25, agent-production-support #26–#27), and list_projects shows current status fine. So writes work; the cross-project log query is wrong (likely ordering, a stale cache/view, or a filter excluding newer projects/sources). The dashboard's recent-activity feed probably uses it, so it looks like no updates since July. Also: vajra-mobile-app notes #22–#24 contain stray '</note></invoke>' fragments; sanitize on write and clean those rows.

claude-project-tracker · claude-project-tracker · by saravanan@scrumclaw.ai

Commit 58dd2aaf2 adds taxonomyGatePublicErrorNote, collapsing any note starting "Tool failure" or "Tool execution failed" to the bare string "Tool failed", in both the on-screen display and the downloaded HTML export. Hiding internal tool names from a customer-facing report is defensible. The problem is that it is applied to the only artifact QA uses for diagnosis. Every root cause in the consolidated fix plan was read out of that exact string: jscpd-cs tool_not_ready:MISSING_BINARY, semgrep tool_not_ready:FAILED_STARTUP_CHECK:VERSION_MISMATCH, dotnet_restore_failed:, lizard VERSION_MISMATCH. It also cuts against fix-plan items 4.1 and 4.2, which ask for MORE provenance per row (tool name, artifact path, raw value), not less. Recommendation: keep the detail behind an internal or admin view, or export an internal variant alongside the customer one. Do not let the only artifact lose it. This is a product decision, not an engineering one — make it deliberately rather than shipping it by default.

Testable (Pramora) · testable-platform · by saravanan@scrumclaw.ai

697e0dd47 raised the declared semgrep version to 1.176.0 for BOTH csharp and python. 76032c488 then made readiness permit a mismatch only when observed >= declared. So any cell still serving an image built with semgrep 1.70.0 or 1.95.0 will now be refused on both languages — where before, Python worked. The image-digest staleness check in 76032c488 should force reconciliation, but that needs verifying rather than assuming. A C#-only smoke test will not catch this. Run a Python-language repository and confirm semgrep still produces output. Symptom if a cell does not reconcile: semgrep reporting VERSION_MISMATCH on a language that worked the day before.

Testable (Pramora) · testable-platform · by saravanan@scrumclaw.ai

Every one of the 12 runs reviewed reports Coverlet, altcover, nyc, vitest, Stryker.NET and StrykerJS producing nothing. That is the entire "are the tests any good?" half of the product, dark in every run. Do NOT accept the claim that the capability is missing. coverlet_runner.py:159-162 runs `dotnet test --collect:"XPlat Code Coverage"`, and altcover_runner.py, stryker_net_runner.py, stryker_ts_runner.py and vitest_coverage_runner.py all exist. Adding a new harness would duplicate working code. (One uploaded consolidated report claimed the opposite and cited runtime_planner.py:120, which is a function that caps planner loop iterations.) Steps: take one run where these rows read "planned tasks did not start", pull the planner's decision for those specific tool assignments, and establish which of three things happened — the task was never created, it was created and never leased, or it was dropped at the readiness gate. Each has a different fix. Only then write the ticket. This is fix-plan item 1.6 and is the only item where the cause is genuinely unknown.

Testable (Pramora) · testable-platform · by saravanan@scrumclaw.ai

Inference runs every 5 min (Scheduler prediction_calls → ml /run_prediction) on 120 DARNN models in gs://nlp_ml_files/prediction_files last written 2025-10-30. Training lives in CoinTrends_Python (local Windows GPU box, localhost Flask chain :17000 → :5000). Target: Cloud Run Job or Vertex AI custom job with GPU, parameterised by asset + minute offset, triggered by Scheduler, creds from Secret Manager, metadata write-back to cryptopredictionslicemaster.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

ModelCreation_latest.py calls cursor.commit() — psycopg2 cursors have no commit() (needs cnxn.commit()), and the second NaN branch uses cnxn which is only defined in the first branch. Result: a failed training run never gets marked activeflag='I'. Also .cuda() hard-coded; pyodbc import leftover.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

core-service, ml-core-service and web-service all connect as postgres superuser on a shared instance. Create e.g. cointrends_core (rw on CryptoTrack_Prod.dbo), cointrends_ml (read + prediction tables), cointrends_web (CryptoUserBilling_Prod), cointrends_ro (SELECT only, for analysis/Claude).

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

insertDataIntoDB and other handlers call log.Fatalf on upstream/DB errors, which kills the whole Cloud Run instance (CODE_REVIEW #7, #17). Return errors and let the handler respond 500.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Each coin has its own table family (pricedata, ohlcv, predictiondata, indicators, indicators_plus + view). Coin list is hard-coded in CASE statements in getcryptodata, getpricedataforcurrencies, getcurrencydetails, and in crypto_id == 1 or ... lists in ml-core-service and CoinTrends_Python. Design an asset-agnostic schema (asset master + long-format tables keyed by asset_id) and drive code from cryptomaster/asset master.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Take *_latest.py versions only. Fold master_create_update_latest.py decision logic (create vs incremental update, check_record_count) and api_latest.py/ModelCreation_latest.py training into one CLI entrypoint: train --asset BTCUSD --minute 15. Remove Flask/localhost HTTP, hardcoded DB host/password, SA key, Windows paths; read config from env. Fix cursor.commit() and NaN-branch bugs (#91). Make device selection automatic (cuda if available else cpu). Once secret-free, this can go into a git repo (unblocks decision #108).

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Dockerfile on a CUDA-enabled PyTorch base, pinned versions, psycopg2 + cloud-sql-python-connector + google-cloud-storage. Push to us-central1-docker.pkg.dev/prj-crypto538-prod-260529/crypto538-container-images/ml-training. Cloud Build config alongside.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

gcloud run jobs create ml-training --gpu 1 --gpu-type nvidia-l4 (us-central1; confirm GPU quota/region availability first). Dedicated SA with only: Cloud SQL Client, read/write on gs://nlp_ml_files/prediction_files, secret accessor on the DB secret. --tasks = assets × 12 offsets (120 today), --parallelism capped to quota; CLOUD_RUN_TASK_INDEX → (asset, minute). Codify in crypto538-terraform. Note: #98 — only this SA should be able to write model files.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Write new models to a staging prefix first, backtest vs the Oct-2025 models (MAE/RMSE, directional accuracy) on recent data, then promote by updating cryptopredictionslicemaster paths. Keep the old files as rollback.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Batch training currently connects as `postgres`, the admin of the shared Cloud SQL instance crypto-track-321e8b45, which other products also use. Before scheduling, create a user with access to only the CryptoTrack_Prod objects training needs: getcryptodata, cryptomaster, cryptopredictionslicemaster and updatetracking. Store its password in a new secret and point batch/*.json at it.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Three approaches were considered: (a) install the generic github/github-mcp-server, (b) extend this plugin, (c) build a custom MCP between the two. Chose (b) because the plugin already runs git commit/push at session_end — adding "comment on linked PR when a todo completes" and "auto-close items when PR with `Closes #<item_id>` merges" sits naturally there. Phased rollout: install GitHub MCP for ad-hoc access first, then plugin frontend badges, then webhook handler, then optional issue mirror.

claude-project-tracker · claude-project-tracker · by saravanan@scrumclaw.ai

Pure frontend change in the Go templates. When a ref looks like a GitHub URL (github.com/.../pull/N or .../issues/N), render it as a styled badge with the PR/issue number. Other refs render as code chips like today. Templates touched: detail.html, pending.html, possibly dashboard.html log lines.

claude-project-tracker · claude-project-tracker · by saravanan@scrumclaw.ai

In tracker_complete_item, if the item's refs[] contains a GitHub PR URL and a GITHUB_TOKEN is configured, POST a comment to the PR like "✅ tracker item #&lt;id&gt; ('&lt;title&gt;') marked done by &lt;user&gt;". Best-effort — log on failure, don't block the tracker update. Requires the plugin to know the GitHub token (new optional env var or read from gh CLI config).

claude-project-tracker · claude-project-tracker · by saravanan@scrumclaw.ai

These BRD v3.0 acceptance items are code-complete but marked Ready for QA / Pending QA — manual walkthroughs and the prod p50 latency measurement were deferred until ship.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

RoadIntelix · roadintelix · by Prajith

Consolidated tech-debt item (full register: docs/TECH_DEBT.md). Verified against the repo after a whitebox gate report (run 61e7d93e, commit 11639ae) whose own scores were unreliable. Sev1 (test enforcement): backend 51 pytest files but conditional-skip → no-op in CI (seed DB + key/mock); frontend+mobile have 0 unit tests; tsc --noEmit + lint not required PR checks (TS6133 noise). Sev2 (silent failures): bare except:pass in exam_service.py:175/194 + note_service.py:278; insights_service.py:479 most_improved/needs_attention hard-coded None (parent digest degrades). Sev3 (gates): no pre-commit secret scanning (gitleaks), no npm audit/pip-audit in CI. Sev4 (cleanup): delete mobile/App.tsx.full duplicate; remove dead comingSoon/*_ENABLED branches. Sequence Sev1→4; each ships as a small PR. No feature impact.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

Web has frontend/src/app/components/PassageHighlighter.tsx (Bluebook-style select-to-highlight + per-highlight notes, char-offset ranges). Mobile ExamSessionScreen renders question_text as a plain <Text> (line ~503) with no highlight/annotate. Gap noticed by owner while testing the full exam on mobile. Non-trivial in RN: no DOM selection; needs a custom passage renderer (tappable/selectable spans → highlight ranges), persisted per-question highlights, optional notes, erase/clear. NOT an App Store blocker — sequence AFTER the build-7 resubmission clears review. Ref: mobile/src/screens/ExamSessionScreen.tsx, frontend/src/app/components/PassageHighlighter.tsx.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

Backend constant minTopupAmount is 1 but the error message and frontend both say the minimum is 100. Change minTopupAmount to 100 so the backend actually enforces what the message claims — otherwise a bespoke client can top up ₹1 (bypasses the intended floor). - const minTopupAmount = 1 + const minTopupAmount = 100

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Frontend passes `me.phone_number || storedEmail || ''` into Razorpay's prefill.contact. That field expects a phone number and is used to auto-detect UPI apps. Passing an email makes checkout look wrong and skips UPI autofill. Fix: only set prefill.contact when we have an actual phone number; put the email into prefill.email instead. prefill: { name: userName, contact: me?.phone_number?.trim() || undefined, email: storedEmail || undefined, }

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

The webhook endpoint currently accepts any POST from anywhere. HMAC signature check is the only defense — cryptographically sound, but each invalid request still triggers signature compute + JSON parse + DB lookup, so it's a cheap DoS vector. Two things worth adding (nginx-level is easiest): 1. Restrict source IPs to Razorpay's published webhook IPs (they publish a list): allow only those in the nginx server block for /webhooks/*. 2. Rate-limit /webhooks/* at nginx (limit_req_zone). Even a modest 20 rps burst limit stops volumetric noise. Doesn't change functionality, just hardens.

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

BRD Q8 spec locked: - A5 landscape (5.8" × 8.3"), vinyl, laminated for outdoor durability - Contents: Razorpay QR image (large, centered) + human-readable station+gun ID (e.g., "Vajra Volt · VAJRA0001 · Gun 2") + typical-cost hint ("Typical 30-min charge ≈ ₹150–300") - Support phone number visible - English only (Q20 answer) - Print files: SVG + PDF at 300 DPI - Vendor: existing signage vendor already identified (Q17) Sample layout: [ Vajra Volt logo ] [ QR CODE 4"x4" ] VAJRA0001 · Gun 2 Scan any UPI app to pay Typical charge ≈ ₹150–300 Support: 88831-61155 Deliverable: printable design ready for vendor.</body>

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Behind WALKIN_QR_SMS_ENABLED flag. 5 templates × 4 languages (see #53).\n\nBackend changes:\n- Templates stored in Go source (map[templateKey]map[lang]string) with DLT template IDs from #53\n- Language selection: prefer user.preferred_language if set; fallback to English\n- Dispatch service reuses existing Fast2SMS integration (already in codebase for OTP)\n- Fire-and-forget from webhook handler (async goroutine); log failures but don't block webhook response\n\nBlocked on DLT approval (#53).</body>

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Useful for ops debugging (missing webhook, dispute lookup, reconciliation). Wraps Razorpay: GET https://api.razorpay.com/v1/payments/qr_codes/:qr_id/payments Backend endpoint: GET /admin/chargers/:charger_id/connectors/:connector_id/payments Returns array of payments captured against that gun's QR (all-time), each with vpa, contact, email, amount, timestamp, status. Enables ops to answer "who paid at this gun today" without querying our DB. Small addition to the admin endpoints in #65. Roughly 1 hour of Go + admin console UI work.</body>

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Chose to mount walkin's routes on the main API router instead of deploying cmd/walkin-service on port 8082. Reason: Saravanan did not want two processes to supervise on a 2GB VM, and the "microservice" boundary was not real — same DB, same .env, same host, same webhook secret. Commit c17a33a, merged to app as 3f826d7. cmd/walkin-service deleted in b3b0fa7 with docs updated. Trade-off accepted: forwardWalkinQRCredit in wallet_handler.go still makes an HTTP round-trip to this same process (WALKIN_SERVICE_URL and API_BASE_URL both = http://127.0.0.1:8080). Both are deliberately loopback, NOT https://api.vajraev.com — routing back out through nginx during webhook handling can tie up nginx workers waiting on a process that is waiting on nginx. Cleanup owed: collapse the self-call into a direct service-method call and drop both env vars. Deferred because it touches the payment path. Whoever wrote cmd/walkin-service has NOT been told it was removed — worth asking whether there was a reason for the split that the docs never recorded.

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

47b5c3a1a worked around fix-plan item 1.2 by reassigning jscpd-cs to the node family and lizard, semgrep, semgrep-perf-static and pydriller to python — a legitimate choice, executed thoroughly (it also fixed the node alias key cve-lite -> cve-lite-cli and updated csharp_tool_readiness.py to return NOT_APPLICABLE for borrowed tools). But docker/runtime/Dockerfile.dotnet still has no ARG INSTALL_MANIFEST and no COPY of it — confirmed unchanged between origin/qa and origin/dev. Unlike Dockerfile.node:11,24 and Dockerfile.python:15,31, the reconciler generates a manifest for it and the template discards it. Any future tool assigned to the dotnet family will fail the same way, silently. Not a release blocker; close the gap when the dotnet family next needs a tool it does not already ship.

Testable (Pramora) · testable-platform · by saravanan@scrumclaw.ai

The consolidated fix plan says of ESLint: "--eslintrc was removed in v9; remove the flag or pin ESLint 8." That is wrong. The platform already pins ESLint 8 in the node image. The real fault, found by the team in 77f8b9d14 (TP-5958), was binary resolution: `npx eslint` resolved the REPOSITORY's ./node_modules/.bin/eslint — version 9 — which rejects the flag the image pin still requires. Removing the flag, as the plan advised, would have broken the image's ESLint 8 behaviour. Anyone acting on the plan's original wording should stop and read 77f8b9d14 instead. Note this does NOT address fix-plan item 2.3, the missing JSX parser — no ESLint configuration file is touched in the 53 files changed between qa and dev, so React frontends still return "Unexpected token &lt;".

Testable (Pramora) · testable-platform · by saravanan@scrumclaw.ai

The suggested per-kind paths returned 404. Working endpoints: - GET /api/projects/adaptive-sat/items — all statuses - GET /api/items?project_id=adaptive-sat — open items only There are no separate todos/blockers/commitments routes. Every entry is an item, and its `kind` field (todo/blocker/decision/note) says which type it is.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

April 2026 review lists 20 open issues. Highlights: no HTTP client timeouts (#6), unbounded goroutines in fetchCryptoNews (#8), Choices[0] with no bounds check (#3), double db.Close() in UpdateCryptoNewsExtractByURL (#4), T-SQL @ID syntax in IsNewsExtractNull fails on Postgres (#5), no MaxBytesReader (#11), rows.Err() unchecked (#12).

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

getDbConnection() → connectWithConnector() builds a new sql.DB (and Cloud SQL dialer) on nearly every call, with defer db.Close(). Create one pool at startup and reuse it; set pool limits once.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

crypto538-webapp uses REACT_APP_COIN_DATA_BASE_URL pointing at the public core-service Cloud Run URL, bypassing web-service auth/plan checks. Needed before core-service can be made private.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

ml-core-service torch.load(..., weights_only=False) and np.load(allow_pickle=True) on files from GCS: anyone who can write to that bucket can execute code in the ML service. Limit bucket writers to the training job identity; consider safetensors / weights-only state_dict.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

One DARNN per 5-minute offset (desired_minute 0..55) → 12 models per coin, 120 today. Consider one model per asset with minute-of-hour as a feature, and a horizon suited to each asset class (e.g. daily for stocks/indexes).

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

crypto538-askgenie-service (Ollama llama3, 16 GB, 4 CPU, min-instances=1, no CPU throttling — billed 24/7) plus a separate llama-cpu Cloud Run service plus OpenAI calls from core. Pick one backend for Genie/recommendations and scale the rest to zero or delete.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

/pricing promises "Live chat with an Analyst" and "On-demand automated trading agent" (not built). Premium lists the same 10-coin access as Standard. Either build, mark "coming soon", or remove.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Razorpay was only used for the VajraEV experiment. Code kept for now with deprecation comments (PR #70); deletion later.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

The folder holds the DB password, public DB IP and an SA key in every script. Revisit after the secret-rotation items are done.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Instance (POSTGRES_14, 1 vCPU / 3.75 GB, 20 GB) is shared with AdaptiveSAT, Vajra, CitrineOS and the tracker. Schema digests in the claude.ai CoinTrends project: claude/cointrends-db-schema-notes.md and claude/cointrends-architecture-notes.md.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Cloud sessions started from the Cowork folder have no GitHub repo access, so Claude edits the local checkout and Saravanan runs ship.sh (branch → commit → push → PR → squash-merge → delete branch → ff base → git fetch --prune). No orphaned branches.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Chosen over Vertex AI custom jobs: same platform as the rest of the stack (Cloud Run, Artifact Registry, Cloud Build, Scheduler), pay-per-execution, task-index fan-out fits the asset × minute-offset model grid. Parent work item: #90.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Cloud Scheduler → Cloud Run Jobs run API (cadence TBD, e.g. weekly full / daily incremental using the existing updatetracking logic). Alert on failed executions and on stale models (modelfile older than N days). Retire the Windows box flow.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Fresh (5-min): pricedata (~518k rows/coin since 2024-10-29), indicators, marketindex, metalsdata, predictiondata, pricedatacache. Stopped: OHLCV Mar/Apr 2025; tradingrecommendation 2025-03-26; cryptonews & feargreedindex 2025-10-03; gainers/losers/trending 2025-10-13; updatetracking last 2025-01-30 (so the retrain baseline is 20 months old); TON price 2026-06-17. Empty: predictionvsactual, predictiontracking. Slice master has 24 slices per coin (5..120 min) all flagged active with modelfile set, but only slice 24 (120 min) models exist in GCS and are used. Legacy usertable 8 rows, paymentdetails 5 rows.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Saravanan 2026-10-19: do not stop or change anything that is running (schedulers, Cloud Run services, DB rows such as slices 1-23, live models). Findings and fixes are collected now; all production changes happen together at the end as an ordered cleanup plan where every step has a validation check and a rollback.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Only production runs in the cloud. Leftovers: 3 idle staging Cloud Run services in prj-crypto538-stage-443113 (no triggers deploy to them), stage.cointrends.ai LB commented out in terraform, CryptoTrack_Stage / CryptoUserBilling_Stage DBs on the prod instance. Merging to the `stage` branch deploys nothing; `master` deploys prod. Leftovers can be reviewed in the final cleanup.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Saravanan 2026-10-20: train on GPU using Google Cloud Batch. Shape: same ml-training container (CUDA build) in Artifact Registry; Batch job with one task per asset (BATCH_TASK_INDEX), g2-standard-4 (1x L4) Spot VMs, installGpuDrivers, parallelism ≤ quota (L4: 8 on-demand / 8 preemptible in us-central1), dedicated service account, DB via Cloud SQL connector, password from Secret Manager, logs to Cloud Logging. Triggered by Cloud Scheduler calling the Batch jobs API. Batch API not yet enabled. Updates #113/#114/#116.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Deferred until items #4 and #5 (the close-on-merge webhook and the PR badges) prove valuable. When ready: a background goroutine in the Go server polls `GET /repos/:owner/:repo/issues` every N minutes and upserts each issue as a tracker item (linking via parent_item_id or a new `external_id` column). Two-way write-back on completion. Per-project repo config in claude_projects (probably a new repo_url is already there).

claude-project-tracker · claude-project-tracker · by saravanan@scrumclaw.ai

The claude-collab session captured 1 decision + 4 todos on the GitHub integration, but they were stamped with project_id='outputs' because the plugin's cwd in that session was Cowork's outputs folder, not the project folder. The v3.4 ephemeral-folder guard prevents this for tracker_init_project but doesn't retroactively fix existing rows. Cleanup options: (a) DELETE FROM claude_projects WHERE project_id='outputs' CASCADE in Studio, accepting the loss of those originals (they're now duplicated under claude-project-tracker by this digest), or (b) UPDATE the 5 rows to set project_id='claude-project-tracker'.

claude-project-tracker · claude-project-tracker · by saravanan@scrumclaw.ai

All 7 Ask AI Tutor requirements are intentionally Deferred (decision D6, 2026-05-17). v3.0 scope is 13 enhancements (#9–#22 minus #20), 108 in-scope REQs. Tracking as a note so it isn't mistaken for unplanned work.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

Per decision on 2026-06-13, iPad support is not required for the initial App Store submission (iPhone-only). Revisit in a future phase: iPad layout/adaptivity pass + iPad screenshots in App Store Connect. No target date.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

Razorpay deprecated the payment_capture parameter for the Orders API — capture behavior is now controlled at the account/merchant level. Not a functional bug today, but noise, and Razorpay may eventually reject unknown fields in future API versions. Drop the key from the payload in createRazorpayOrder.

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Where to find each customer field in the qr_code.credited webhook payload. Confirmed from Razorpay official docs (docs/webhooks/qr-codes). **Payload path for VPA:** `payload.payment.entity.vpa` **Trimmed sample:** ```json { "event": "qr_code.credited", "contains": ["payment", "qr_code"], "payload": { "payment": { "entity": { "id": "pay_HO2fEpc9JeOQU5", "amount": 200, "method": "upi", "vpa": "gauri.kumar@okhdfcbank", "email": "gauri.kumari@example.com", "contact": "+919000090000", "customer_id": "cust_HKsR5se84c5LTO", "notes": {}, "acquirer_data": { "rrn": "116812981837" } } }, "qr_code": { "entity": { "id": "qr_HO2e0813YlchUn", "notes": { "charger_id": "VAJRA0001", "connector_id": "2" } } } } } ``` **Two subtle gotchas:** 1. **Two separate `notes` blocks** — `payload.payment.entity.notes` is per-payment (usually empty for QR), and `payload.qr_code.entity.notes` is the QR's notes (where we set charger_id + connector_id at QR creation). Read the second one for gun identification. 2. **VPA presence:** always present for QR-triggered payments (method is always upi). Fallback: null. **Go struct extraction:** see TDD §4.3 `qrCreditedPayload`. **Fallback retrieval (webhook lost):** - `GET /v1/payments/pay_XXX` — fetch single payment; includes vpa - `GET /v1/payments/qr_codes/qr_XXX/payments` — list all payments for a QR (great for admin ops "show all payments to this gun") Sources: - https://razorpay.com/docs/webhooks/qr-codes/ - https://razorpay.com/docs/webhooks/payloads/payments/</body>

Vajra Volt Mobile App · vajra-mobile-app · by saravanan@scrumclaw.ai

Appended TRACKER_PROJECT_ID=adaptive-sat to the env file. TRACKER_USER was already saravanan@scrumclaw.ai, so it was left alone. No other lines changed; file permissions are still 600.

Adaptive SAT · adaptive-sat · by saravanan@scrumclaw.ai

Delete razorpay_webhook.go, razorpay_webhook_test.go and the /webhooks/razorpay route; also the dead /users and /payments handlers. Zero traffic in 30 days of Cloud Run logs. Deprecation comments merged 2026-10-16 (c3ef045).

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Duplicate public.getcryptodata variants, public.gainers_json, public.market_metal_data, dbo.table_1, dbo.table_2, dbo.debugtable, dbo.jsondatatowork; legacy dbo.usertable / dbo.paymentdetails / manageuserandpayments() (users now live in CryptoUserBilling_Prod). Check row counts and archive before dropping.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Test step commented out in core-service cloudbuild.yaml; stage LB and purge job commented out in terraform; webapp/web-service/terraform sit on a stage branch diverged from master; READMEs empty; log retention 30 days. Also split the 3.9k-line core main.go into packages.

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai

Footprint so far (none of it changes how prod runs): (1) IAM: Cloud SQL service account p475046183936-kjdopc@gcp-sa-cloud-sql granted objectAdmin on gs://crypto538-prj-crypto538-prod-260529-bucket (needed for schema/query exports) — rollback: gsutil iam ch -d. (2) gs://crypto538-prj-crypto538-prod-260529-bucket/claude-schema-exports/ (2 schema-only .sql files; query CSVs are deleted after each read). (3) Cloud Build 7d835ecf may have pushed ml-training:test-cpu / :test-gpu to Artifact Registry crypto538-container-images — delete if unused. (4) No Cloud Run jobs, no DB writes, no scheduler changes. Local only: ml-training/, ship.sh, commit_razorpay_deprecation.sh, 'Tracker project_id' line in cointrends/CLAUDE.md. Merged code change: PR #70 (comments only).

CoinTrends.ai · cointrends · by saravanan@scrumclaw.ai