Skip to content

CHANGELOG.md

Latest commit

 

History

History
150 lines (113 loc) · 10.1 KB

CHANGELOG.md

File metadata and controls

150 lines (113 loc) · 10.1 KB

Change Log

All notable changes to the "github-copilot-api-vscode" extension will be documented in this file.

Check Keep a Changelog for recommendations on how to structure this file.

[2.14.0] - 2026-05-18

Added

  • Comprehensive Telemetry: Expanded from 7 sparse events to 15 rich telemetry helpers, giving deep observability via Azure Application Insights.
    • Activation context: VS Code version, OS platform/arch, Node version, Copilot Chat presence, extension version.
    • Deactivation summary: Session duration bucket, total commands and requests served this session.
    • Server start (extended): WebSocket enabled, MCP enabled, model family, max concurrency, API key presence, rate-limit presence.
    • Server stop (extended): Uptime bucket, total requests, error rate bucket.
    • Request details (extended): Endpoint bucket, token-in/out buckets, message-count bucket, streaming flag, tool use flag, system-prompt flag, finish reason.
    • New request.error event: Per-request error category (auth / rate_limit / timeout / server_error / client_error).
    • New rateLimit.hit event: Endpoint and limit type when a 429 is returned.
    • New model.switched event: Old/new model family and trigger source (quickpick / api / config).
    • New config.changed event: Setting key changed (never the value) and whether a server restart followed.
    • New ws.* events: WebSocket connected, message (type only), disconnected, error lifecycle.
    • New uriHandler.invoked event: Deep-link path used (/dashboard, /start, /stop).
    • Performance heartbeat (perf.heartbeat): Fires every 5 minutes while the server is running — heap MB bucket, uptime, RPM, error rate, server state.
    • All 3 chat commands now tracked: openCopilotChat, askCopilot, askSelectionWithCopilot — previously had no telemetry.
    • Shared bucketing utilities: durationBucket, tokenBucket, messageCountBucket, uptimeBucket, heapMbBucket, modelFamily exported from TelemetryService to eliminate duplicated logic.
    • bucketEndpoint() helper: Maps raw URL paths to stable, low-cardinality App Insights dimension labels.

Changed

  • telemetryServerStarted props extended (hostBucket, enableWebSocket, enableMcp, modelFamily, maxConcurrency, hasApiKey, hasRateLimit).
  • telemetryServerStopped now includes uptime and request-count context.
  • telemetryServerError and telemetryTunnelStarted updated to use structured props objects.

Dependencies

  • lint-staged 17.0.2 → 17.0.4
  • typescript-eslint 8.59.2 → 8.59.3
  • @types/node 25.6.0 → 25.7.0
  • ws 8.20.0 → 8.20.1 (security fix: uninitialized memory disclosure in websocket.close())

[2.10.2] - 2026-03-09

Fixed

  • Release workflow: explicitly package the VSIX before uploading to GitHub Release, ensuring the release page actually contains the artifact.

[2.10.1] - 2026-03-09

Fixed

  • Release packaging: removed stale .vsix files from repository root so published release contains only the current version.

[2.10.0] - 2026-03-09

Added

  • Claude Code Full Compatibility: End-to-end support for Claude Code connecting via the Anthropic Messages API (/v1/messages).
  • x-api-key Authentication: Now accepts Anthropic SDK-style x-api-key header in addition to Authorization: Bearer — required for Claude Code with API key auth enabled.
  • Tool Use in Anthropic Path: The streaming Anthropic handler now passes tools to the VS Code LM API and emits proper tool_use SSE content blocks (content_block_start → input_json_delta → content_block_stop) with stop_reason: "tool_use".
  • system Prompt as Array: Both streaming and non-streaming Anthropic handlers now accept system as an array of text blocks ([{type:"text",text:"..."}]) per the latest Anthropic SDK spec.
  • AnthropicContentBlock Type: Added explicit TypeScript union type for text, tool_use, and tool_result content blocks.
  • Extended AnthropicMessageRequest: Interface now includes tools, tool_choice, and array system fields.

Fixed

  • Claude 500 Error — "Unexpected chat message content type llm 2": Multi-part array message content ([{type:"text",text:"..."}]) was incorrectly JSON.stringify'd before being passed to the VS Code LM API. Now uses flattenMessageContent() across all three conversion paths (processStreamingChatCompletion, invokeCopilotWithTools, processStreamingAnthropicMessages).
  • tool_result Content Blocks: flattenMessageContent() now correctly extracts text from tool_result blocks in message history, allowing multi-turn tool conversations.
  • tool_use Content Blocks: Conversation history messages containing tool_use blocks are now formatted as human-readable summaries instead of silently dropped.
  • CORS Headers: Added x-api-key, anthropic-version, and anthropic-beta to Access-Control-Allow-Headers for full Anthropic SDK CORS compatibility.

[2.9.0] - 2026-02-20

Added

  • Modern Dashboard UI: Complete visual overhaul with glassmorphism-inspired cards, refined color palettes adapting to VS Code themes, and smooth interactive hover effects.
  • Multi-Model Support Badge: Highlighted capability in the dashboard hero section noting the Gateway can fetch ANY language model detected in VS Code.
  • Wiki Refinements: Redesigned documentation tabs to a modern pill-shaped style and improved code block typography.

Fixed

  • Live Feed & Activities: Resolved JavaScript reference errors that broke the real-time Live Log and Recent Activity features.
  • Cloudflare Tunnel Status: Fixed a bug where starting a tunnel displayed the literal string "null" instead of a proper "Starting" state in the dashboard.

[2.7.0] - 2026-02-15

Added

  • Status Bar: Shows active model name, uptime counter, tunnel indicator (🌐), and enriched tooltip with full metrics table
  • Quick Pick Menu: Copy API URL, Quick Test (sends live "Hello" request), Switch Model (lists all Copilot models), Edit System Prompt, Start/Manage Tunnel
  • Sidebar: Animated pulsing status dot, clickable model name for switching, live uptime ticker (ticks every second), 4-stat grid (RPM, latency, total reqs, error rate), live request feed with flash animation, config status indicators (Auth/HTTPS/Tunnel)
  • README: "Run as a Background Service" guide for macOS, Windows, and Linux

Changed

  • Sidebar layout reorganized: stats and live feed now above action buttons for better at-a-glance monitoring
  • Removed duplicate Swagger button event listener from sidebar

[2.6.0] - 2026-02-13

Added

  • OpenAI Chat Completions: max_completion_tokens parameter support (auto-normalized to max_tokens for GPT-5.x compatibility)
  • OpenAI Chat Completions: developer role support (auto-mapped to system for 2025+ spec)
  • OpenAI Chat Completions: stream_options.include_usage — emit usage chunk in streaming responses
  • OpenAI Chat Completions: reasoning_effort parameter support for o-series models (auto-mapped to reasoning.effort)
  • OpenAI Responses API: text.format structured output pass-through (was hardcoded to text)
  • OpenAI Responses API: truncation parameter pass-through
  • OpenAI Responses API: Expanded reasoning.effort values — added minimal, none, xhigh (Aug–Dec 2025 spec)
  • OpenAI Responses API: Streaming events now pass through text.format and truncation from request
  • Anthropic Messages API: thinking, metadata interface fields; tool_use stop reason; cache token usage
  • Google Generative AI: frequencyPenalty, presencePenalty, responseMimeType, responseSchema, safetySettings interface
  • OpenAPI Spec: Added max_completion_tokens, reasoning_effort to Chat Completions schema
  • OpenAPI Spec: Added developer to Message role enum
  • OpenAPI Spec: Added reasoning, truncation, store, previous_response_id, tool_choice to Responses API schema

Changed

  • Updated branding to emphasize free, open-source, and trustworthy nature
  • Upgraded TypeScript to 5.9.3, typescript-eslint to 8.55.0, @types/node to 25.2.3

[2.5.1] - 2026-02-07

Fixed

  • Cloudflare Tunnel binary: Now downloads the cloudflared binary automatically at runtime to extension storage. Works properly for marketplace-installed extensions.

[2.5.0] - 2026-02-07

Added

  • 🌐 Internet Access via Cloudflare Tunnels: Expose your API to the internet with a single click. Get a free public *.trycloudflare.com URL instantly — no account required. Perfect for accessing from your phone, sharing with friends, or remote development.
  • Network Access Guide: New "What's New" banner explaining the difference between localhost (127.0.0.1), LAN (0.0.0.0), and Cloudflare Tunnel access.
  • Security Enforcement: Tunnel requires API key authentication to be enabled before going live.

Changed

  • Dashboard UI improvements with better feature discovery.

[2.1.5] - 2026-01-11

Fixed

  • Real-time metrics: Sidebar and dashboard now update metrics (Req/Min, Latency, Tokens, Connections) in real-time
  • Dashboard race condition: Fixed message listener timing issue that caused stale data on initial load
  • Host/Port layout: Fixed overlap between host, port inputs and Apply button

Added

  • "Things you should read" button in sidebar linking to notes.suhaib.in

[2.1.4] - 2026-01-10

Fixed

  • Model Selection: Fixed critical issue where API endpoints were ignoring the requested model and defaulting to the first available model. Now all endpoints strictly validate and use the exact model specified in the request.
  • Added model validation across all API endpoints (OpenAI, Anthropic, Google, Llama). Invalid models now return a 404 error with a list of available models.

Changed

  • Updated README to be model-agnostic and expanded tool categories.
  • Removed specific model name references to future-proof documentation.

[0.0.7] - 2025-12-21

Fixed

  • Dashboard readability issues in Light and High Contrast themes by using VS Code theme variables.

[0.0.6]

  • Initial release