Add support for AWS Bedrock, Azure OpenAI, OpenRouter, and new features

.github/workflows/ci.yml

@@ -0,0 +1,29 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+jobs:
+  test:
+    name: test (${{ matrix.os }})
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, macos-latest, windows-latest]
+
+    steps:
+      - uses: actions/checkout@v4
+      - uses: pnpm/action-setup@v4
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 20
+          cache: pnpm
+      - run: pnpm install --frozen-lockfile
+      - run: pnpm run release:check-assets
+      - run: pnpm test
+      - run: pnpm run compile:cli
+      - run: pnpm run benchmark:smoke
+      - run: pnpm run eval:smoke

.github/workflows/release.yml

@@ -0,0 +1,28 @@
+name: Release
+
+on:
+  push:
+    tags:
+      - 'v*'
+
+jobs:
+  package:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+    steps:
+      - uses: actions/checkout@v4
+      - uses: pnpm/action-setup@v4
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 20
+          cache: pnpm
+      - run: pnpm install --frozen-lockfile
+      - run: pnpm run compile
+      - run: pnpm exec vsce package
+      - name: Generate release notes
+        run: node dist/cli.js changelog > release-notes.md
+      - uses: softprops/action-gh-release@v2
+        with:
+          body_path: release-notes.md
+          files: '*.vsix'

.gitignore

@@ -8,4 +8,24 @@ dist/
 .thunder V2/
 .mitii
 project-goals/
-mitii-debug-logs/
+mitii-debug-logs/
+
+# Benchmark / eval runtime (local generation and reports)
+tools/benchmark/tasks/eval/generated/
+tools/benchmark/tasks/eval/generated-smoke/
+tools/benchmark/results/
+
+# Fixture sandboxes — deps, builds, Mitii session state
+tools/benchmark/fixtures/**/.mitii/
+tools/benchmark/fixtures/**/.next/
+tools/benchmark/fixtures/**/dist/
+tools/benchmark/fixtures/**/node_modules/
+tools/benchmark/fixtures/**/package-lock.json
+tools/benchmark/fixtures/**/pnpm-lock.yaml
+
+# Agent-created during eval (not part of pinned fixture baseline)
+tools/benchmark/fixtures/**/docs/
+tools/benchmark/fixtures/**/index.html
+tools/benchmark/fixtures/**/src/index.css
+tools/benchmark/fixtures/**/src/main.ts
+tools/benchmark/fixtures/**/src/main.tsx

.npmrc

@@ -0,0 +1,3 @@
+# vsce runs `npm list` to resolve production dependencies; pnpm's default
+# isolated linker hoists transitive packages that npm treats as extraneous.
+node-linker=hoisted

.vscode/tasks.json

@@ -11,7 +11,7 @@
     {
       "label": "thunder: rebuild native",
       "type": "shell",
-      "command": "npm run rebuild:native",
+      "command": "pnpm run rebuild:native",
       "problemMatcher": [],
       "presentation": { "reveal": "always", "panel": "shared" }
     },

.vscodeignore

@@ -18,6 +18,7 @@ scripts/**
 .gitignore
 .vscodeignore
 package-lock.json
+pnpm-lock.yaml
 tsconfig.json
 vite.config.ts
 vitest.config.ts
@@ -26,4 +27,5 @@ vitest.config.ts
 
 !dist/**
 !media/**
-!bundled-skills/**
+
+tools/benchmark/**

CHANGELOG.md

@@ -0,0 +1,31 @@
+# Changelog
+
+## [Unreleased]
+
+## [2.7.17] - 2026-07-02
+
+### Added
+- Implemented the backlog end-to-end across the repo. What changed: Added diff-first micro-task routing for commit messages, changelog entries, and release notes. Unified SCM commit message generation through the new micro-task executor. Added changelog/release generation core, VS Code commands, CHANGELOG.md, and CI release workflow. Added one-click audit pack export with zip generation, redaction report, manifest, tool audit, and approvals. Improved reasoning stream UI: live reveal, no 1200-char hard cap, configurable visibility/preview size. Added enterprise settings/docs for local-only providers, audit redaction, procurement/security/compliance. Added Windows path hardening, Windows CI matrix, and Windows smoke checklist. Added CLI MVP for changelog, prepare-release, and export-audit. Updated README with the new enterprise, audit, release, CLI, and reasoning features. Added focused tests for all new core implementations. (8a00fd537ef2)
+- add AWS Bedrock, Azure OpenAI, and OpenRouter provider support (cac73bc1a9a7)
+- **modes:** add pilot and enterprise depth levels to Ask, Plan, and Act modes (ca3d2daf6375)
+- **orchestrator:** add planning clarification question flow with resume support (0d052937fb8e)
+- auto-discover verify commands, add retry with install, cap sequential-thinking calls (6786a1d904fe)
+- improve plan step UI, add skipped tool handling, and strip channel markers (d65a81752037)
+- add GitHub token setting and improve issue comment pagination (7c8e53d68036)
+- Implemented the core folder structure migration safely. (017ddd4d0766)
+- auto-fetch GitHub issue context when user pastes an issue URL (1f0b368dfb50)
+- add retrieval timing metrics, repeated tool failure guard, and Act mode MCP exclusions (dda2e8f38881)
+- add Act orchestration boundary for Agent mode execution (0620f2e55902)
+- **docs:** update README with enhanced project description, features, and usage instructions (920bf12f6f38)
+- **plan:** add skill-aware planning pipeline with phased PlanPanel UI (e23f19d9e929)
+- bundle skill playbooks inside extension and auto-install on workspace init (07a83238eb87)
+- add planDepth setting and skill frontmatter parsing (d77ed8f6062f)
+- **plan-mode:** add Plan mode orchestration with intent routing and read-only grounding (14372af57d12)
+- **core:** add file read caching and parallelize read files tool (95dee1dbc8ac)
+- **scm:** add AI-generated commit messages and ask-mode depth controls (8ea94111e43f)
+- **ask:** add structured ask mode with intent routing, scope resolution, and impact analysis (aa2660f9ef3d)
+
+### Changed
+- add project-goals entry to .gitignore configuration (81ecf0da086c)
+- Merge branch 'main' of https://github.com/codewithshinde/thunder-ai-agent (957469705bc4)
+- **tools:** remove unused resolveToolDirPath and related imports (44a3dd3d0a8c)

CONTRIBUTING.md

@@ -20,7 +20,7 @@ For bugs and feature ideas, open an [issue](https://github.com/codewithshinde/th
 |------|---------|
 | VS Code | 1.85+ (or Cursor, with native rebuild noted below) |
 | Node.js | 20+ |
-| npm | 9+ (ships with Node 20) |
+| pnpm | 10.13+ |
 | git | any recent version |
 
 Optional but useful for full feature coverage:
@@ -36,12 +36,12 @@ Optional but useful for full feature coverage:
 ```bash
 git clone https://github.com/codewithshinde/thunder-ai-agent.git
 cd thunder-ai-agent
-npm install
-npm run rebuild:native   # required for better-sqlite3 in VS Code
-npm run compile
+pnpm install
+pnpm run rebuild:native   # required for better-sqlite3 in VS Code
+pnpm run compile
 ```
 
-Git hooks are installed automatically via `npm install` → `prepare` → `scripts/install-git-hooks.mjs`. The pre-commit hook stages version bumps from `scripts/bump-version.mjs`.
+Git hooks are installed automatically via `pnpm install` -> `prepare` -> `scripts/install-git-hooks.mjs`. The pre-commit hook stages version bumps from `scripts/bump-version.mjs`.
 
 ### Launch the extension
 
@@ -53,7 +53,7 @@ Git hooks are installed automatically via `npm install` → `prepare` → `scrip
 ### Watch mode (day-to-day dev)
 
 ```bash
-npm run watch
+pnpm run watch
 ```
 
 Rebuilds the extension bundle and webview on save. Reload the Extension Development Host window after extension-side changes (`Ctrl/Cmd+R` in the host window, or restart the debug session).
@@ -63,7 +63,7 @@ Rebuilds the extension bundle and webview on save. Reload the Extension Developm
 ## Project layout
 
 ```
-thunder-ai-agent/
+mitii-ai-agent/
 ├── src/
 │   ├── extension.ts              # VS Code entry point
 │   ├── core/                     # Agent logic (editor-agnostic)
@@ -89,7 +89,9 @@ thunder-ai-agent/
 │       └── src/
 ├── test/                         # Vitest tests
 ├── scripts/                      # Build, audit, hook helpers
+├── tools/benchmark/              # @mitii/benchmark — fixtures, enterprise + eval harness
 ├── dist/                         # Compiled output (gitignored)
+├── pnpm-workspace.yaml           # Workspace: tools/*
 └── package.json                  # Extension manifest + settings schema
 ```
 
@@ -100,7 +102,20 @@ thunder-ai-agent/
 
 Scaffolds may exist at `mitii-docs/` and `mitii-website/` in this tree while you split them out. Brand constants: `src/shared/brand.ts` (sync with each repo's `brand.ts`).
 
-**Rule of thumb:** keep VS Code APIs out of `src/core/`. Core should be testable without launching an editor. Put platform glue in `src/vscode/`.
+**Rule of thumb:** keep VS Code APIs out of `src/core/`. Core should be testable without launching an editor. Put platform glue in `src/vscode/`. Keep benchmark and eval in `tools/benchmark/` — they are not extension runtime code.
+
+### Benchmark and eval
+
+```bash
+pnpm run compile:cli
+pnpm run benchmark:smoke
+pnpm run eval:preflight       # before real-runtime eval
+pnpm run eval:generate
+pnpm run eval:standard -- --provider openai-compatible \
+  --base-url http://localhost:11434/v1 --model qwen3-coder:30b --limit 50
+```
+
+See [tools/benchmark/README.md](tools/benchmark/README.md) for sharded runs, Ollama matrix, and Inspect AI.
 
 ---
 
@@ -109,23 +124,23 @@ Scaffolds may exist at `mitii-docs/` and `mitii-website/` in this tree while you
 ### Run tests
 
 ```bash
-npm run rebuild:node   # if better-sqlite3 fails under vitest
-npm test               # full suite
-npm run test:watch     # watch mode
-npm run smoke          # smoke tests only
+pnpm run rebuild:node   # if better-sqlite3 fails under vitest
+pnpm test               # full suite
+pnpm run test:watch     # watch mode
+pnpm run smoke          # smoke tests only
 ```
 
 ### Typecheck
 
 ```bash
-npm run lint           # tsc --noEmit
+pnpm run lint           # tsc --noEmit
 ```
 
 ### Build a VSIX
 
 ```bash
-npm run compile
-npm run package        # outputs thunder-ai-agent-<version>.vsix
+pnpm run compile
+pnpm run package        # outputs thunder-ai-agent-<version>.vsix
 ```
 
 Install locally: **Extensions → ... → Install from VSIX**.
@@ -134,20 +149,20 @@ Install locally: **Extensions → ... → Install from VSIX**.
 
 | Scenario | Command |
 |----------|---------|
-| F5 / VS Code extension host | `npm run rebuild:native` |
-| Cursor extension host | `THUNDER_EDITOR=cursor npm run rebuild:native` |
-| Local vitest | `npm run rebuild:node` |
-| Both | `npm run rebuild:all` |
+| F5 / VS Code extension host | `pnpm run rebuild:native` |
+| Cursor extension host | `THUNDER_EDITOR=cursor pnpm run rebuild:native` |
+| Local vitest | `pnpm run rebuild:node` |
+| Both | `pnpm run rebuild:all` |
 
 If SQLite throws on startup, this is almost always the fix.
 
 ### Audit scripts
 
 ```bash
-npm run audit:dependencies
-npm run audit:dead-code
-npm run check:circular-deps
-npm run audit:engines
+pnpm run audit:dependencies
+pnpm run audit:dead-code
+pnpm run check:circular-deps
+pnpm run audit:engines
 ```
 
 These are useful before large refactors. Not required on every PR, but run them if you touch imports or dependencies.
@@ -174,7 +189,7 @@ The pre-commit hook may stage a version bump in `package.json`. Include that in
 
 ### Code style
 
-- TypeScript strict mode — `npm run lint` must pass
+- TypeScript strict mode - `pnpm run lint` must pass
 - Match surrounding patterns: no drive-by refactors in unrelated files
 - Logging via `createLogger('ComponentName')` from `src/core/telemetry/Logger.ts`, not raw `console.log`
 - New VS Code settings go in `package.json` contributes **and** `src/core/config/schema.ts` **and** `src/core/config/vscodeSettings.ts`
@@ -200,7 +215,7 @@ The pre-commit hook may stage a version bump in `package.json`. Include that in
 
 1. Fork and branch from `main`
 2. Make your change; keep the diff focused
-3. Run `npm run lint` and `npm test`
+3. Run `pnpm run lint` and `pnpm test`
 4. Manually smoke-test in the Extension Development Host if you touched agent behavior or UI
 5. Open a PR against `main` with:
    - What changed and why (2–4 sentences is fine)