CHANGELOG.md

Changelog

[Unreleased]

Added

• OpenAI compatibility: emit standard CORS headers on /v1/responses for non-streaming requests, and return 400 on unsupported non-streaming usage.
• Deploy-time SW versioning: gen:sw-version writes /sw-version.json; registrar registers /sw.js?v=… so each deploy gets an isolated cache namespace.
• SW shell-cache + deploy-regression test seams (tests/unit/swShellCache.test.js, tests/SW-TEST-SEAM.md).

Changed

• bun run check / lint gate on oxlint --deny-warnings.

Fixed

• SW navigation: network-first (not cache-first) with offline /offline fallback; never Response.error() on documents/images; drop blind controllerchange reload (f1d4861).
• Remove redundant controller.close() in open-sse/handlers/chatCore.js finally block — already closed in the success path.
• Canary body-size latency: extend readBodyTextStream() chunk-by-chunk reads across additional large-body routes to avoid the 9–15s curl/8.x stall.
• Redis rate-limit isolation: respect RATELIMIT_KEY_PREFIX so environments (e.g. canary/prod) keep separate namespaces.
• Client disconnect: classify AbortError at node:_http_server as [ClientDisconnect] (not [FATAL]) and controller.close() on SSE reader abort — no unhandled rejections, no log spam.
• Body size cap raised to 50MB default, env-tunable via POD_MAX_REQUEST_BODY_BYTES and POD_MAX_CHAT_BODY_BYTES.
• /api/monitoring/health* are public reads (auth guard removed); older changelog “+ API key auth” entries are historical only.

v0.0.82 (2026-07-11)

Fixed

• Canary 9-15s body-size-bound latency: src/sse/handlers/chat.ts now uses readBodyTextStream() (chunk-by-chunk body read with explicit size cap) instead of request.text() for large bodies. Added per-request timing instrumentation (t_read, t_parse, t_bypass) for diagnosis.
• Redis shared-keyspace risk: src/lib/rateLimit/redis.ts now respects RATELIMIT_KEY_PREFIX env var so canary can use a separate namespace. Added 1-second per-op timeout wrapper (RATELIMIT_REDIS_TIMEOUT_MS env) to prevent stalled connections from blocking requests.
• Stale semantic cache doc: .agents/architecture/{03-data,01-app}.md now correctly describe the cache as in-process LRU + per-instance SQLite (no Redis involvement), matching src/lib/semanticCache.ts.

v0.0.81 (2026-07-11)

Fixed

• AbortError unhandledRejection spam on client disconnect: request.text() / request.json() now return { ok: false, reason: "aborted" } via readBodyText() helper; routes return 499 cleanly. SSE stream wrappers in src/sse/handlers/chat.ts use controller.close() (not controller.error(err)) on reader abort, preventing the abort from re-surfacing in the Next.js response writer.
• Hard 10MB body cap raised to 50MB default (env-tunable: POD_MAX_REQUEST_BODY_BYTES, POD_MAX_CHAT_BODY_BYTES).
• Global unhandledRejection / uncaughtException handlers in server-init.ts and instrumentation.ts now classify Error: aborted at node:_http_server as [ClientDisconnect] (not [FATAL]) and dedupe log spam (1s window, 5-error threshold).
• Abort-safe body parsing applied to 6 sibling routes: imageGeneration, tts, fetch, embeddings, search, pricing/sync.
• open-sse/handlers/chatCore.js: catch block now filters AbortError and calls controller.close() instead of controller.error(e).
• Resolve intermittent hang: fix HEALTHCHECK port, SSE counter leak, Redis leak, instrumentation.ts init order, dead code cleanup.
• Redirect root / to /endpoint instead of nonexistent /dashboard.
• Don't block server startup on Redis init failure in register().
• Cache tailscale binary lookup; show cached API keys immediately.
• Add offline cache to combos page — instant load on revisit.
• Optimize service worker for Zeabur cold-start.
• Resolve remaining eqeqeq lint warnings across src/.

v0.0.80 (2026-07-07)

Docs

• Add API compatibility matrix (compatibility-matrix.md) with OpenAI/Anthropic endpoint status, error shapes, and SDK verification.
• Add API compatibility policy and reference-checking workflow to AGENTS.md.
• Update README and AGENTS.md for regex fix and project scope.

Fixed

• Update stale JS/no-TS references in architecture docs.
• Escape unterminated regex literal in logger.ts for Turbopack.
• Set minimumLockoutMinutes=0 in test seeds to avoid 60m default skew.
• Anthropic error format compliance: catch blocks now return Anthropic error shape on /v1/messages.
• Add missing error code fields in route handlers.

v0.0.79 (2026-06-05)

• Security hardening phase 1: error sanitization across 18+ API routes (sanitizeError)
• Security hardening phase 2: safe JSON body parsing across 45+ routes (parseJsonBody)
• Security hardening phase 3: upstream API body leak patches (13 additional routes)
• Security hardening phase 4: SSE/streaming crash hardening (3 containment points)
• Redis rate limiting with automatic backend selection
• Duck-type backend dispatch (replace constructor.name/instanceof — breaks in minified builds)
• RPM slot release when concurrent check fails after RPM passes
• Variable shadowing fixes in 3 OAuth routes

v0.0.78 (2026-06-04)

• Fix Codex tool call response cut-off + reasoning effort normalization

v0.0.77 (2026-06-03)

• Security hardening + upstream adoption + race fixes

v0.0.76 (2026-06-02)

• Fix CHANGELOG, node:fs imports, .env.example branding

v0.0.75 (2026-05-29)

• Fix Codex reasoning token budget + audit findings
• feat: Account Lockout Status section on /health page
• feat: 'Connected only' toggle on /media-providers/web page
• feat: connection-level lockdown with exponential cooldown

v0.0.67 (2026-05-27)

• Fix Web Cookie providers showing 'No connections' + filter respect

v0.0.66 (2026-05-27)

• Verify combo systemPrompt works for OpenCode Go (both routes)

v0.0.65 (2026-05-26)

• Lock-in combo system prompt injection across all 6 provider shapes

v0.0.64 (2026-05-26)

• /health page works for dashboard users + fix metrics

v0.0.63 (2026-05-26)

• Enrich /api/monitoring/health with cache/sync/runtime metrics + API key auth

v0.0.62 (2026-05-26)

• Clean noisy console.log spam in client components

v0.0.61 (2026-05-26)

• Fix /api/usage/stats 400 on period=90d

v0.0.60 (2026-05-26)

• Silence "no label associated with a form field" warnings — aria-label

v0.0.59 (2026-05-26)

• Silence form-field autofill warnings — add id/name/aria-label

v0.0.58 (2026-05-26)

• Silence Chrome 'preloaded but not used' CSS warning

v0.0.57 (2026-05-26)

• Fix Add OpenAI/Anthropic Compatible modal — 400 silent failure

v0.0.56 (2026-05-24)

• Relay function honours x-relay-timeout + Kiro transient retry

v0.0.55 (2026-05-24)

• Harden Vercel relay (timeout race, platform 504, retry, healthcheck)

v0.0.54 (2026-05-24)

• Perplexity-web perceived TTFT — x-pod-skip-reasoning header

v0.0.53 (2026-05-24)

• Render Web Cookie Providers section in /providers

v0.0.52 (2026-05-24)

• Remove paid Perplexity API provider

v0.0.51 (2026-05-24)

• Harden CodeQL #32-#35 fixes (re-scan still flagged)

v0.0.50 (2026-05-24)

• Resolve all 14 open CodeQL alerts

v0.0.48 (2026-05-23)

• Sink-level log sanitizer for CodeQL #39

v0.0.42 (2026-05-20)

• Fix /usage Details 502 error, replace datetime-local with DatePicker, fix Est. Cost to 2dp
• Fix OAuth: hardcode Codex redirect_uri to localhost:1455, remove dead mitm exports

v0.0.36 (2026-05-19)

• Fix tunnel enable 'Unable to connect' error — 3 root causes fixed
• Rename 'System Health' to 'Health' + align page spacing with /memory and /cache
• Persist ReactFlow viewport to sessionStorage + rename Usage title
• Rename 'Providers' to 'LLM Providers' in header, breadcrumbs, page title
• Center-align 'No connected providers' placeholder in /media-providers

v0.0.33 (2026-05-19)

• Fix /quota double-click expand bug + add Last Request At column
• Fix /usage Details tab always showing empty — observability config + id propagation

v0.0.32 (2026-05-18)

• Fix Codex 502 invalid JSON response, remove disableCodexStreaming workaround

v0.0.31 (2026-05-18)

• Add memory pipeline, retrieval, store integration tests (711 tests)
• Fix clearInFlight called in all cache paths, add memory+cache edge case tests (653 tests)
• Fix SQLite TTL datetime format, clearInFlight on large responses, add cache integration tests
• Include memoryOwnerId in cache signature, fix temperature default, bump cache size to 512KB

v0.0.30 (2026-05-17)

• Add 26 Vertex AI stream guard tests
• Add missing testClaude stub, fix cloud branding, add 39 cloud tests
• Sanitize browser network errors in tunnel enable, wrap fetchData non-fatal, remove DNS warmup delay

v0.0.29 (2026-05-16)

• Add INITIAL_PASSWORD, BASE_URL, CLOUD_URL env vars to README

v0.0.28 (2026-05-16)

• Restore hideDisabled toggle behavior in /quota, remove permanent filter
• Ping tunnelUrl instead of publicUrl for faster tunnel health check
• Remove requestTooLargeForCache guard (generateSignature handles large payloads)
• Skip stream field injection for Vertex AI target format
• Fix models.dev sync — read cost from .cost field, fix 0 model count
• Remove Melma from API Key providers

v0.0.27 (2026-05-16)

• Replace 9Router branding with Pod across repo
• Add .pi to gitignore

v0.0.25 (2026-05-15)

• Fix recheckAndClear re-lock respects backoff multiplier
• Migrate Inter and IBM Plex Mono to next/font, suppress Material Symbols lint

v0.0.24 (2026-05-15)

• Add perf-cpu-hotpath tests for SSE hotpath fixes
• Reduce CPU load from SSE hot paths

v0.0.23 (2026-05-15)

• Resolve 14 CodeQL security alerts
• Fix: skeleton on initial load across health/usage/cache/memory/endpoint pages
• Fix: model lockout recheck before clear, spinner icon, human-readable duration
• Fix: console logs "Showing x of y logs" below log list
• Fix: combo test uses /api/models/test endpoint, spinner icon
• Feat: quota toolbar toggles with session persistence + logs live state

v0.0.22 (2026-05-14)

• Fix provider topology active node border white instead of orange
• Fix providers/media-providers Connected Only filter + remove arrow buttons
• Fix semantic cache always 0% hit rate — temperature/top_p normalization

v0.0.21 (2026-05-14)

• Feat: health page SSE live update
• Fix minimum lockout time not applied correctly
• Persist Connected Only toggle state in localStorage

v0.0.20 (2026-05-14)

• Resolve remaining 14 CodeQL security alerts
• Update RWX workflows for pod repo

v0.0.19 (2026-05-14)

• Feat: configurable minimum lockout time with exponential multiplier
• Fix semantic cache 0% hit rate due to memory injection signature mismatch
• Fix console/request logs height
• Fix quota page toolbar improvements
• Health page improvements

v0.0.18 (2026-05-14)

• Exclude disabled models from provider quota progress
• Resolve all 23 CodeQL security alerts

v0.0.17 (2026-05-13)

• Feat: drag-to-reorder combos and provider connections
• Multi-account custom providers

v0.0.15 (2026-05-13)

• Fix Codex provider freeze during tool calls
• Add CONTRIBUTING.md, SECURITY.md, update README

v0.0.14 (2026-05-13)

• Add paseo.json to .gitignore, remove melma-router binary

v0.0.13 (2026-05-13)

• Fix memory leaks — reduce RSS from 1.2GB to ~200-400MB
• Fix browser tab title "Pod ✦ Pod ✦ Health" → "Pod ✦ Health"
• Fix semantic cache 0% hit rate — temperature threshold too strict
• Bound usage_history growth, debounce stats SSE

v0.0.12 (2026-05-13)

• Add bun run check script — format, lint, type check

v0.0.11 (2026-05-13)

• Fix usage chart — linear line instead of curved
• Request logs SSE: detect PENDING→SUCCESS/FAILED, poll 1s when pending
• Fix usage stats Cost/Tokens pill tabs — standardize to SegmentedControl

v0.0.9 (2026-05-13)

• Fix usage chart white graph color for Tokens/Cost
• Fix console logs timestamp column width
• Improve upstream error logging — URL, model, status code
• Model test bypasses semantic cache via x-pod-no-cache header
• Request log detail: abort previous fetch on new selection

v0.0.8 (2026-05-13)

• Memory extraction: add Indonesian language patterns for preference/decision/habit

v0.0.7 (2026-05-13)

• Thundering herd protection for semantic cache via in-flight deduplication
• LRU connection cache, cooldown guard, queue depth metrics, rate limit docs

v0.0.6 (2026-05-13)

• Security: API key auth on GET /v1/models endpoints; timing-safe comparison
• Kebab-case URL redirects for web-search/web-fetch
• Move Connected Only toggle to header action slot
• Provider nodes: optional Identifier field
• Adopt upstream fixes — developer role norm, stream stall timeout, Ollama usage, Gemini schema, DATA_DIR fallback, Today period, Blackbox provider, MiniMax TTS

v0.0.5 (2026-05-12)

• Standardize Button and Badge usage across app
• Quota page: Collapse All, rename Progress→Quota column
• Console logs: Refresh/Live toggle, Connecting status fix
• Combos page: Test button per combo
• Provider health grouped by provider, show provider icon
• Remove MITM bypass feature entirely
• Settings: remove Migrate to SQLite button

v0.0.4 (2026-05-12)

• Replace all browser confirm() with in-app ConfirmModal across all pages
• Fix breadcrumb label alignment
• Fix memory page crash — missing confirmDialog state

v0.0.3 (2026-05-12)

• Console logs: toolbar in header, level filter pills
• Request logs: search/filter, SSE live stream, detail payload fix
• Proxy logs: silent refresh, configured count always visible
• Standardize pill tabs with SegmentedControl across /usage and /logs
• Quota page: grouped by provider with 3-level expand/collapse
• Remove close button from sonner toasts
• Add @custom-variant dark for Tailwind v4 dark mode

v0.0.2 (2026-05-11)

• Health page: Rate Limit Status, Model Lockout Status, sparklines
• Logs page: multi-tab (Request Logs, Proxy Logs, Console)
• Usage page: ReactFlow topology, metric cards redesign, Details tab
• Replace custom toast with sonner, position bottom-right
• Theme toggle, light theme implementation
• API keys table with pagination (15 per page)
• Sidebar: larger app name, version label, custom SVG logo

v0.0.1 (2026-05-10)

• Fork from 9Router — reset version to v0.0.1
• Rebrand: 9Router → Pod across code, branding, Docker image, data dir
• Migrate runtime from Node.js to bun
• Remove /dashboard prefix from all routes
• Replace better-sqlite3 with bun:sqlite (SQLite via bun native)
• Docker image: lazuardytech/pod
• CI/CD: GitHub Actions build + Docker publish on tag push
• OpenSSE imported locally (jsconfig.json alias)
• Core pages adapted: login, endpoint, providers, logs, usage, health, settings, memory, cache