Enhance memory management, CLI functionality, and team services

.github/workflows/mitii-agent.yml

@@ -0,0 +1,28 @@
+name: mitii agent
+
+on:
+  issue_comment:
+    types: [created]
+
+jobs:
+  mitii:
+    if: contains(github.event.comment.body, '/mitii')
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+      pull-requests: write
+      issues: 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:cli
+      - run: node dist/cli.js agent "${{ github.event.comment.body }}" --provider echo --approval auto --json > mitii-events.jsonl
+      - uses: actions/upload-artifact@v4
+        with:
+          name: mitii-audit
+          path: mitii-events.jsonl

.github/workflows/native-binaries.yml

@@ -0,0 +1,38 @@
+name: native binaries
+
+on:
+  push:
+    tags:
+      - 'v*'
+  workflow_dispatch:
+
+jobs:
+  build:
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - os: macos-14
+            target: darwin-arm64
+          - os: macos-13
+            target: darwin-x64
+          - os: ubuntu-latest
+            target: linux-x64
+          - os: windows-latest
+            target: win32-x64
+    runs-on: ${{ matrix.os }}
+    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 --filter mitii build:platform
+        env:
+          MITII_PLATFORM_TARGET: ${{ matrix.target }}
+      - uses: actions/upload-artifact@v4
+        with:
+          name: mitii-${{ matrix.target }}
+          path: dist-native/${{ matrix.target }}

.github/workflows/npm-publish.yml

@@ -0,0 +1,32 @@
+name: npm publish
+
+on:
+  push:
+    tags:
+      - 'v*'
+  workflow_dispatch:
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      id-token: write
+    steps:
+      - uses: actions/checkout@v4
+      - uses: pnpm/action-setup@v4
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 20
+          registry-url: https://registry.npmjs.org
+          cache: pnpm
+      - run: pnpm install --frozen-lockfile
+      - run: pnpm sync:versions
+      - run: pnpm -r build
+      - run: pnpm test
+      - run: npm publish --workspace packages/sdk --access public
+        env:
+          NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}
+      - run: npm publish --workspace packages/daemon --access public
+        env:
+          NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}

.github/workflows/release.yml

@@ -19,7 +19,7 @@ jobs:
           cache: pnpm
       - run: pnpm install --frozen-lockfile
       - run: pnpm run compile
-      - run: pnpm exec vsce package
+      - run: pnpm run package
       - name: Generate release notes
         run: node dist/cli.js changelog > release-notes.md
       - uses: softprops/action-gh-release@v2

.gitignore

@@ -7,8 +7,17 @@ dist/
 .thunder/
 .thunder V2/
 .mitii
+.mitii/
+.mitti/
 project-goals/
 mitii-debug-logs/
+vendor/
+tmp/
+logs/
+build/
+coverage/
+*.lock
+*.map
 
 # Benchmark / eval runtime (local generation and reports)
 tools/benchmark/tasks/eval/generated/

.mitiiignore

@@ -0,0 +1,11 @@
+# Mitii workspace indexing ignores
+node_modules/
+dist/
+build/
+.next/
+coverage/
+vendor/
+tmp/
+logs/
+*.lock
+*.map

docs/deploy/launchd.plist

@@ -0,0 +1,22 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+  <dict>
+    <key>Label</key>
+    <string>dev.mitii.serve</string>
+    <key>ProgramArguments</key>
+    <array>
+      <string>/usr/bin/env</string>
+      <string>mitii</string>
+      <string>serve</string>
+      <string>--hostname</string>
+      <string>127.0.0.1</string>
+      <string>--port</string>
+      <string>4310</string>
+    </array>
+    <key>RunAtLoad</key>
+    <true/>
+    <key>KeepAlive</key>
+    <true/>
+  </dict>
+</plist>

docs/deploy/systemd-user.service

@@ -0,0 +1,11 @@
+[Unit]
+Description=Mitii daemon
+
+[Service]
+Type=simple
+EnvironmentFile=%h/.mitii-serve-env
+ExecStart=/usr/bin/env mitii serve --cwd %h/project --hostname 127.0.0.1 --port 4310
+Restart=on-failure
+
+[Install]
+WantedBy=default.target

docs/developers/mitii-serve-protocol.md

@@ -0,0 +1,75 @@
+# Mitii Serve Protocol
+
+`mitii serve` runs one daemon per bound workspace. It exposes a loopback-first HTTP API plus per-session SSE event streams.
+
+Default bind: `127.0.0.1:4310`.
+
+## Security
+
+- Loopback is the default.
+- Binding to `0.0.0.0` or another non-loopback host requires `--insecure-bind`.
+- Non-loopback binds also require `--token` or `MITII_SERVER_TOKEN`.
+- Clients authenticate with `Authorization: Bearer <token>`.
+- CORS is denied unless `--allow-origin` is set.
+- Daemon actions are appended to `.mitii/daemon/audit.jsonl`.
+
+## Routes
+
+`GET /health`
+
+```json
+{ "ok": true, "version": "2.7.31", "cwd": "/repo", "sessions": 1 }
+```
+
+`GET /capabilities`
+
+```json
+{
+  "features": ["sessions", "sse", "permissions", "cancel", "subagents", "worktrees"],
+  "maxSessions": 5,
+  "supportedModes": ["ask", "plan", "agent", "review"],
+  "eventReplay": true
+}
+```
+
+`POST /session`
+
+```json
+{ "cwd": "/repo", "mode": "agent", "approval": "manual", "runtime": "real" }
+```
+
+`GET /sessions`, `GET /session/:id`, and `DELETE /session/:id` list, inspect, and close sessions.
+
+`POST /session/:id/prompt`
+
+```json
+{ "mode": "agent", "message": "Fix the failing test", "attachments": [] }
+```
+
+Returns `202` once the prompt is accepted. Results stream through SSE.
+
+`GET /session/:id/events`
+
+SSE frames use `id`, `event`, and JSON `data` fields. Reconnect with `Last-Event-ID` to replay buffered missed events.
+
+```text
+id: 7
+event: assistant_delta
+data: {"type":"assistant_delta","content":"I found the issue"}
+```
+
+`POST /session/:id/permissions/:approvalId/respond`
+
+```json
+{ "decision": "approved" }
+```
+
+`POST /session/:id/cancel` aborts an in-flight turn.
+
+## Error Shape
+
+```json
+{ "error": { "code": "workspace_mismatch", "message": "Daemon is bound to ..." } }
+```
+
+Common statuses: `400` workspace mismatch, `401` auth failure, `404` missing session, `409` concurrent prompt, `503` session limit.

docs/enterprise/PROCUREMENT_FAQ.md

@@ -10,18 +10,18 @@ Mitii targets macOS, Linux, and Windows. CI runs `npm test` on `ubuntu-latest`,
 
 ## Offline And Local Models
 
-Mitii supports OpenAI-compatible localhost providers such as Ollama and LM Studio. Set `thunder.enterprise.localProvidersOnly` to enforce local-only usage.
+Mitii supports OpenAI-compatible localhost providers such as Ollama and LM Studio. Set `mitii.enterprise.localProvidersOnly` to enforce local-only usage.
 
 ## Controls Procurement Teams Ask For
 
 | Control | Setting or command |
 |---|---|
-| Disable session logging | `thunder.telemetry.sessionLogging` |
-| Disable verbose diagnostics | `thunder.telemetry.debugMetrics` |
-| Require approval for writes | `thunder.safety.requireApprovalForWrites` |
-| Require approval for shell | `thunder.safety.requireApprovalForShell` |
-| Local providers only | `thunder.enterprise.localProvidersOnly` |
-| Strip file contents from audit packs | `thunder.enterprise.stripFileContentsFromAuditPacks` |
+| Disable session logging | `mitii.telemetry.sessionLogging` |
+| Disable verbose diagnostics | `mitii.telemetry.debugMetrics` |
+| Require approval for writes | `mitii.safety.requireApprovalForWrites` |
+| Require approval for shell | `mitii.safety.requireApprovalForShell` |
+| Local providers only | `mitii.enterprise.localProvidersOnly` |
+| Strip file contents from audit packs | `mitii.enterprise.stripFileContentsFromAuditPacks` |
 | Export review evidence | `Mitii: Export Audit Pack` |
 
 ## License

docs/enterprise/README.md

@@ -7,5 +7,5 @@ This folder is the reviewer-facing security, compliance, and procurement packet
 - [Procurement FAQ](PROCUREMENT_FAQ.md)
 - [Enterprise overview](MITII_ENTERPRISE_OVERVIEW.md)
 
-Key controls are implemented as VS Code settings under `thunder.telemetry.*`, `thunder.safety.*`, `thunder.enterprise.*`, and the `Mitii: Export Audit Pack` command.
+Key controls are implemented as VS Code settings under `mitii.telemetry.*`, `mitii.safety.*`, `mitii.enterprise.*`, and the `Mitii: Export Audit Pack` command.
 

docs/enterprise/SECURITY.md

@@ -4,15 +4,15 @@ Mitii is local-first. Workspace indexes, memory, plans, checkpoints, and JSONL s
 
 ## What Leaves The Machine
 
-Only prompt data sent to the configured LLM provider leaves the machine. The provider boundary is controlled by `thunder.provider.*`. Enterprises can enable `thunder.enterprise.localProvidersOnly` to require Echo or a localhost OpenAI-compatible endpoint.
+Only prompt data sent to the configured LLM provider leaves the machine. The provider boundary is controlled by `mitii.provider.*`. Enterprises can enable `mitii.enterprise.localProvidersOnly` to require Echo or a localhost OpenAI-compatible endpoint.
 
 ## Secrets
 
-API keys are stored in VS Code SecretStorage. Logs and audit packs redact keys named like API keys, tokens, passwords, and secrets. Audit packs can additionally strip tool output and file content with `thunder.enterprise.stripFileContentsFromAuditPacks`.
+API keys are stored in VS Code SecretStorage. Logs and audit packs redact keys named like API keys, tokens, passwords, and secrets. Audit packs can additionally strip tool output and file content with `mitii.enterprise.stripFileContentsFromAuditPacks`.
 
 ## MCP Risk Model
 
-MCP tools are routed through Mitii tool policy. File writes and mutating shell commands require approval according to `thunder.safety.approvalMode` and `thunder.safety.requireApprovalForWrites`.
+MCP tools are routed through Mitii tool policy. File writes and mutating shell commands require approval according to `mitii.safety.approvalMode` and `mitii.safety.requireApprovalForWrites`.
 
 ## Auditability
 

docs/enterprise/channel-security.md

@@ -0,0 +1,15 @@
+# Channel Security
+
+Channel connectors route Slack, Telegram, Discord, and similar messages into the daemon. Treat them as external ingress.
+
+Baseline controls:
+
+| Control | Mitii behavior |
+| --- | --- |
+| User allowlist | Connectors can reject users not present in `allowedUsers`. |
+| Thread allowlist | Connectors can restrict use to approved channel/thread IDs. |
+| Read-only mode | Channel policy can downgrade `agent` requests to `plan`. |
+| Secret redaction | Long token-like values are redacted from outbound replies. |
+| Auditability | Channel prompts flow through daemon sessions and session logs. |
+
+Enterprise deployments should keep connectors bound to loopback daemons or require `MITII_SERVER_TOKEN` for non-loopback hosts.

docs/enterprise/daemon-security.md

@@ -0,0 +1,12 @@
+# Daemon Security
+
+`mitii serve` is designed for local development first.
+
+- It binds to `127.0.0.1` by default.
+- Non-loopback binds require both `--insecure-bind` and a bearer token.
+- Bearer tokens are compared with constant-time comparison.
+- CORS is opt-in through `--allow-origin`.
+- Session `cwd` values are canonicalized and must match the daemon-bound workspace.
+- Session lifecycle, prompts, cancellation, and permission responses are logged to `.mitii/daemon/audit.jsonl`.
+
+For enterprise-managed installs, prefer a per-user daemon, managed token provisioning, local provider policy enforcement, and workspace-level audit pack export.

docs/enterprise/managed-policy.md

@@ -0,0 +1,27 @@
+# Managed Enterprise Policy
+
+Mitii can be governed with a managed `.mitii/mitii.policy.json` file distributed by MDM, base images, or repository templates. Policy values are intentionally coarse and override user-facing convenience features before a connector, worker, or team runtime starts.
+
+```json
+{
+  "localProvidersOnly": true,
+  "channelsDisabled": false,
+  "maxParallel": 10,
+  "autoPrEnabled": false,
+  "stripFileContentsFromAuditPacks": true,
+  "allowedChannelUsers": [],
+  "allowedChannelThreads": []
+}
+```
+
+Recommended controls:
+
+| Policy | Purpose |
+| --- | --- |
+| `localProvidersOnly` | Require local or approved private model providers. |
+| `channelsDisabled` | Disable Slack, Telegram, Discord, and similar external surfaces. |
+| `maxParallel` | Cap parallel sessions, workers, and teammates. |
+| `autoPrEnabled` | Keep PR creation explicit unless centrally approved. |
+| `stripFileContentsFromAuditPacks` | Preserve audit metadata while reducing data exposure. |
+
+Phase 3 treats this as the stable policy contract. Future SSO/OIDC and HA daemon work should extend this file instead of introducing a second policy surface.