flowseeker 0.1.7

package/CHANGELOG.md

@@ -0,0 +1,111 @@
+# FlowSeeker Changelog
+
+## 0.1.7 - Update Notice and Deterministic Accuracy
+
+- **Update notice:** Extension checks npm for newer FlowSeeker versions at most once per day, shows a non-modal update notice, supports dismissing a specific version, and fails silently when offline.
+- **MCP CLI version checks:** `flowseeker-mcp --version` prints the package version, and `flowseeker-mcp --check-updates` checks npm explicitly without running the MCP stdio server.
+- **MCP protocol safety:** Normal MCP server startup does not perform passive network update checks and keeps stdout reserved for JSON-RPC only.
+- **Caller/callee candidate expansion:** Workspace index now records extracted call names and uses symbol/call relationships to expand candidate files beyond imports.
+- **Generated/noise downranking:** Generated folders, snapshots, mocks, protobuf/gRPC outputs, bundles, framework build folders, and similar noise paths are downranked without being excluded.
+- **High-precision task terms:** Quoted literals and route-like paths are preserved as literal retrieval terms before broad keyword matching.
+
+### 0.1.7 Files changed
+`src/utils/updateChecker.ts` (new), `src/extension.ts`, `src/cli/mcpServer.ts`, `src/index/workspaceIndex.ts`, `src/pipeline/taskUnderstanding.ts`, `package.json`, `package-lock.json`, `CHANGELOG.md`, `README.md`, `COMPATIBILITY.md`
+
+## 0.1.6 - Accuracy, MCP Native Chat, and Package Readiness
+
+- **Quality Score:** Added single composite metric `computeQualityScore()` combining hit@10, context coverage, avoid-hit rate, MRR, and token reduction into one comparable number (Quality = 0.35·Hit@10 + 0.25·Coverage + 0.20·(1−AvoidHitRate) + 0.10·MRR + 0.10·TokenReduction). Printed in every headless evaluation summary line.
+- **Benchmark expansion:** Wired up 22 fixture repos with 697 labeled cases (including 2 negative cases for AvoidHitRate) via `.flowseeker/eval-benchmark-mega.json`. Added `scripts/validate-eval-paths.js` — deterministic ground-truth checker that mirrors runtime path matching, with `--fix` mode to prune dead paths. Recorded baseline Quality Score at 0.741.
+- **Negative term utilization:** Extended `detectNegativeTerms()` with `except` and `ngoại trừ` (Vietnamese exclusion) cues. Existing ranker penalty (-24 score + contrastContext) already wired via `applyTaskMetadata()`. Tested against the full benchmark; zero regression at the conservative pattern set.
+- **Active file context boost:** Ranker now accepts optional `editorContext` (active file, directory siblings, recently opened tabs) and applies small score boosts (+15 active file, +5 sibling dir, +8 recent tabs). Editor context is purely optional — headless evaluation is completely unaffected.
+- **Deterministic eval validator:** New `scripts/validate-eval-paths.js` validates every expected/avoid/group path resolves to a real fixture file. Mirrors runtime `pathMatches()` semantics (exact, suffix, substring). `--fix` mode prunes dead paths and drops unmeasurable cases.
+- **MCP prompt discovery:** Standalone MCP server now advertises `prompts` capability and supports `prompts/list` + `prompts/get` for `retrieve`, `guide`, and `files`. These prompts map to the stable tools `flowseeker_retrieve`, `flowseeker_guide`, and `flowseeker_files` so Native Chat hosts can show short suggestions when their UI supports MCP prompt/tool autocomplete.
+- **MCP metadata polish:** Tightened tool descriptions and input schema wording for truncated IDE suggestion UIs while keeping the stable tool names unchanged.
+- **Cross-IDE config correctness:** Fixed MCP help/docs to use `mcpServers` for Cursor, Antigravity, Windsurf, Claude Desktop, and `servers` only for VS Code/Copilot `.vscode/mcp.json`.
+- **Antigravity Native Chat note:** Documented current Antigravity behavior: selecting FlowSeeker may insert `@mcp:flowseeker:` without showing second-level command suggestions. FlowSeeker still works through natural language or direct tool names; autocomplete visibility is host-controlled.
+- **npm MCP readiness:** Verified `flowseeker-mcp` builds from `dist/cli/mcpServer.js`, responds to `tools/list` and `prompts/list`, and is included by `npm pack --dry-run` for users who want MCP-only integration without installing the extension.
+
+### 0.1.6 Files changed (vs 0.1.5 baseline)
+`src/pipeline/evaluationMetrics.ts`, `src/cli/runEvaluation.ts`, `scripts/validate-eval-paths.js` (new), `scripts/render-icon.js` (new), `TODO-0.1.6.md` (new), `src/pipeline/taskUnderstanding.ts`, `src/pipeline/ranker.ts`, `src/pipeline/runPipeline.ts`, `src/cli/mcpServer.ts`, `src/mcp/mcpTools.ts`, `notes/Req5.txt`, `notes/mcp-config-paths.md`, `.gitignore`, `.vscodeignore`, `.npmignore` (new), `README.md`, `ARCHITECTURE.md`, `COMPATIBILITY.md`, `MARKETPLACE_DEMO.md`, `UPDATE_GUIDE.md`, `plan.md`
+
+## 0.1.5 - Agent UX Stability, Cross-Host Compatibility, and MCP Server
+
+- **MCP Server:** Added standalone MCP server (`flowseeker-mcp`) exposing FlowSeeker retrieval as MCP tools. Works with Cursor, Antigravity, Windsurf, Claude Desktop, and any MCP-compliant AI client.
+- **MCP tools:** `flowseeker_retrieve` (full Solve Packet), `flowseeker_guide` (agent prompt), `flowseeker_files` (lightweight file list).
+- **Cross-IDE notification:** FlowSeeker now detects Cursor, Antigravity, and Windsurf hosts at startup and shows a one-time guidance notification pointing users to the sidebar chat or MCP setup.
+- **Headless agent prompt:** Added `agentPromptHeadless.ts` — agent solve prompt generator without VS Code dependency, used by MCP server and CLI.
+- **Standalone CLI binary:** `flowseeker-mcp` bin for running the FlowSeeker MCP server outside of any IDE.
+- **Host detection:** FlowSeeker now identifies the editor host at startup (VS Code, Cursor, Antigravity, Windsurf, VSCodium, Theia) and logs the identity + capabilities to the Output channel.
+- **Runtime probing:** Beyond `typeof` checks, FlowSeeker now calls `vscode.lm.selectChatModels()` to count actual available models and vendors — a reliable signal of whether the host exposes usable AI models.
+- **Tier table:** Single source of truth for host capabilities — `native-full` (VS Code/Insiders), `native-unverified` (Cursor, Antigravity, Windsurf — API exists but not confirmed to render), `sidebar-primary` (VSCodium, Theia — no API).
+- **Shared command runner:** Core retrieve→ask flow extracted into `flowCommandRunner.ts` so both native chat and sidebar use the same logic.
+- **Enhanced compatibility check:** `FlowSeeker: Check Host Compatibility` now shows host identity, tier, live model probe results, and offers human-in-the-loop verification for unverified hosts.
+- **Status bar indicator:** Shows current tier mode (`FlowSeeker` for native, `FlowSeeker: Sidebar` for sidebar-only) with tooltip explaining why. Click opens compatibility report.
+- **JSON prompt hardening:** Edit generation instruction moved to END of prompt (after context pack), strengthened wording, added `response_format: { type: "json_object" }` for OpenAI-compatible providers, improved JSON extraction from prose responses.
+- **Per-file apply status:** Edits applied file-by-file with individual success/failure reporting including file name and reason.
+- **Graded provider diagnosis:** `testProvider` classifies errors (401/403 credentials, 429 quota vs rate-limit, 5xx server, network, missing CLI) with actionable fix steps.
+- **MCP server visibility:** Agent policy tooltip shows each MCP server name, type, and status (configured/disabled/misconfigured).
+- **Feedback hardening:** Save confirmation includes file path and replay hint. CLI replay shows original top files for drift detection.
+- Added `COMPATIBILITY.md` with detailed host compatibility matrix and model sharing explanation.
+
+### 0.1.5 Files changed
+`src/runtime/hostDetect.ts` (new), `src/runtime/hostTier.ts` (new), `src/runtime/statusBar.ts` (new), `src/agent/flowCommandRunner.ts` (new), `src/mcp/mcpTools.ts` (new), `src/pipeline/agentPromptHeadless.ts` (new), `src/cli/mcpServer.ts` (new), `src/commands/checkHostCompatibility.ts`, `src/extension.ts`, `src/ui/chatViewProvider.ts`, `src/gateway/aiGateway.ts`, `src/gateway/aiProviders.ts`, `src/gateway/agentPrompts.ts`, `src/gateway/editProposalStore.ts`, `src/cli/runReplay.ts`, `src/pipeline/tokenSavings.ts`, `package.json`, `COMPATIBILITY.md` (new)
+
+## 0.1.4 - Token Proof Panel and Feedback Hardening
+
+- Added sidebar proof panel showing token proof status (complete/partial/unavailable) with differentiated visual states per retrieval card.
+- Added context coverage display in the proof panel showing found/missing required slots.
+- Added proof warnings in the metric card for partial scans, capped evidence, and missing slots.
+- Added one-click "Copy token proof" action that copies a formatted token savings summary including proof status, coverage, and warnings.
+- Feedback save confirmation now includes the saved file path and a replay hint.
+- CLI replay reports now show original top files from feedback snapshots alongside current top files for drift visibility.
+
+## 0.1.3 - Token Savings Proof
+
+- Added measured token savings to FlowSeeker results, comparing Solve Packet tokens against indexed workspace text tokens.
+- Uses local BPE token counting with `cl100k_base` for headless, sidebar, and report metrics, with heuristic fallback only when tokenization fails.
+- Counts the actual prompt through VS Code `LanguageModelChat.countTokens()` when the selected native chat model exposes token counting.
+- Added generic Context Blueprint coverage so FlowSeeker checks required flow slots for each task shape instead of ranking only by keyword score.
+- Updated Solve Packet generation to preserve required context slots, report missing links, and mark likely noise.
+- Improved plan approval safety: responses that only ask to read more files are no longer treated as approvable plans.
+- Added bounded context expansion and retry when the AI explains that the first Solve Packet is insufficient.
+- Added structured-edit handling with diff preview, scope lock, apply actions, and verification command proposals.
+- Added sidebar chat history, provider profiles, model switching, provider model fetch, and compact quota/usage status.
+- Added agent controls for workspace edit approval, safe command proposals, MCP/tool proposals, and user-defined prompt rules.
+- Added host compatibility reporting for Native Chat, Language Model, and MCP support across VS Code-compatible editors.
+- Added sidebar feedback capture and real-world replay support so retrieval misses can become regression cases.
+- Normalized feedback kinds from the sidebar and legacy records so replay metrics classify missing context consistently.
+- Shows compact token reduction in native VS Code Chat after retrieval.
+- Added a sidebar chat fallback for VS Code-compatible editors that do not expose the native VS Code Chat participant API.
+- Added token savings and context coverage metrics to headless and VS Code evaluation reports alongside hit@K, MRR, and scan duration.
+- Keeps token reduction wording technical, with retrieval quality measured separately.
+
+## 0.1.2 - Scan UX and Logo Polish
+
+- Refined the FlowSeeker logo assets for extension icon, activity bar, and native chat usage.
+- Kept chat scan progress compact by updating status lines instead of flooding the conversation.
+- Improved AI plan language behavior so generated plans follow the user's request language when set to auto.
+- Tightened package excludes so Marketplace builds do not ship local caches, source maps, or reference raster assets.
+
+## 0.1.1 - Large Repo Stabilization
+
+- Added persistent workspace indexing and dependency expansion to reduce full-codebase scans.
+- Added scan budgets for max files, bytes, duration, raw evidence, and indexed candidates.
+- Improved Solve Packet output with clearer trace, context buckets, uncertainty notes, and agent guidance.
+- Added headless evaluation support for repeatable retrieval tests across sample workspaces.
+
+## 0.1.0 - Initial Release
+
+- **Core retrieval pipeline**: task understanding, workspace indexing, universal text scan, evidence ranking, evidence graph with framework edges
+- **Solve Packet v1**: structured reasoning context with flow buckets (entry/handler/service/schema/tests), problem hypothesis, noise risk detection, missing link identification, agent instructions
+- **Native VS Code Chat participant**: `@flowseeker /guide`, `/auto`, `/apply` commands with plan-first editing guardrail
+- **Framework adapters**: Laravel (PHP), NestJS (TypeScript), Express (JavaScript/TypeScript) - route->controller, event->listener, controller->service, view rendering, DI injection edges
+- **Structured extraction**: regex-based code structure extraction for 11 languages (TS, JS, Python, PHP, Java, Go, C#, Ruby, Rust, Kotlin) - declarations, calls, string literals, comments
+- **Scope lock**: edits are validated against approved plan scope before applying
+- **Diff preview**: file-level change preview before applying structured edits
+- **Verification loop**: auto-detection of test/lint/typecheck commands after edits, with pass/fail reporting
+- **Headless evaluation**: repeatable retrieval benchmarking with hit@K, MRR, expected-path coverage
+- **VS Code Language Model API integration**: works with the selected VS Code Chat model; AI Gateway fallback
+- **Result sidebar**: ranked evidence tree with checkbox selection, context pack copy, agent prompt copy
+- **Configuration**: `.flowseeker/config.json` workspace config + VS Code settings overrides
+- **Scan budget safety**: max files, max bytes, max time limits prevent runaway scans on large repos

package/COMPATIBILITY.md

@@ -0,0 +1,281 @@
+# FlowSeeker Host Compatibility
+
+FlowSeeker detects host capabilities at runtime via `src/runtime/capabilities.ts`. Every feature that is available on the host **will** be used; features that are missing will gracefully fall back to the sidebar + direct provider path.
+
+Run `FlowSeeker: Check Host Compatibility` for a live report on your current host.
+
+---
+
+## How Model Sharing Works
+
+FlowSeeker uses a **three-tier AI backend** (in priority order):
+
+| Tier | Mechanism | Requires |
+|---|---|---|
+| 1. Direct Provider | API key (OpenAI, Anthropic, Gemini, Ollama, OpenRouter, etc.) | Nothing — works on all hosts |
+| 2. Host Language Model | `vscode.lm.selectChatModels` — uses the model the host already has selected | Host must expose VS Code Language Model API |
+| 3. Custom Gateway | OpenAI-compatible `/v1/chat/completions` endpoint | Nothing — works on all hosts |
+
+When you type `@flowseeker /guide` in **Native Chat**, FlowSeeker tries Tier 1 first (your configured provider key). If no key is configured, it falls back to Tier 2 (the host's own model). If neither is available, it shows actionable buttons to configure a provider.
+
+When you use the **Sidebar Chat**, FlowSeeker uses Tier 1 (direct provider) by default, or Tier 3 (custom gateway).
+
+**Bottom line:** On Cursor, FlowSeeker can use Cursor's models. On any host without Language Model API, you can still connect your own API key.
+
+---
+
+## Compatibility Matrix
+
+| Feature | VS Code | Cursor | VSCodium | Antigravity | Theia / OpenVSX |
+|---|---|---|---|---|---|
+| Sidebar Chat | Yes | Yes | Yes | Yes | Yes |
+| Direct Provider API | Yes | Yes | Yes | Yes | Yes |
+| Native Chat (`@flowseeker`) | Yes | Yes | No | Runtime-detected | No |
+| Host Language Model API | Yes | Yes | No | Runtime-detected | No |
+| MCP Server Provider | Yes | Yes | No | Runtime-detected | No |
+| MCP Tool Invocation | Yes | Yes | No | Runtime-detected | No |
+| Copy Agent Prompt | Yes | Yes | Yes | Yes | Yes |
+| Provider Profiles | Yes | Yes | Yes | Yes | Yes |
+
+- **Yes** = Feature is available and tested on this host.
+- **No** = Host does not expose this API. FlowSeeker degrades to sidebar + direct provider.
+- **Runtime-detected** = Depends on the specific build. Run `FlowSeeker: Check Host Compatibility` to confirm.
+
+---
+
+## How Runtime Detection Works
+
+```typescript
+// src/runtime/capabilities.ts
+hasNativeChat: typeof vscode.chat?.createChatParticipant === "function"
+hasLanguageModels: typeof vscode.lm?.selectChatModels === "function"
+hasMcp: typeof vscode.lm?.registerMcpServerDefinitionProvider === "function"
+```
+
+FlowSeeker checks the actual VS Code API surface at startup. There is no hardcoded host list. If a host ships these APIs, FlowSeeker uses them.
+
+---
+
+## Per-Host Details
+
+### VS Code (Stable / Insiders)
+Toàn bộ tính năng hoạt động. Đây là mục tiêu phát triển chính.
+- Native Chat: `@flowseeker /guide`, `/auto`, `/apply`
+- Language Model API: dùng model đã chọn trong VS Code nếu không có provider key
+- MCP: đầy đủ
+
+### Cursor
+Hoạt động tương đương VS Code. FlowSeeker có thể dùng:
+- Native Chat với `@flowseeker`
+- Model của Cursor thông qua Language Model API (nếu không cấu hình direct provider key)
+- MCP tools
+
+**Lưu ý:** Nếu bạn đã cấu hình API key trong FlowSeeker, nó sẽ dùng key đó trước. Để dùng model của Cursor, để trống phần provider key.
+
+### VSCodium
+Không có Native Chat và Language Model API (đây là các API độc quyền của Microsoft).
+- Sidebar chat + direct provider key hoạt động đầy đủ
+- Copy Agent Prompt cho phép paste context vào bất kỳ AI tool nào
+- MCP không khả dụng
+
+### Antigravity
+Tùy thuộc vào build. Một số build của Antigravity có thể include VS Code Chat API và Language Model API.
+- Chạy `FlowSeeker: Check Host Compatibility` để xem chính xác API nào có sẵn
+- Sidebar + direct provider luôn hoạt động khi extension host hỗ trợ webview/sidebar
+- **MCP config:** `~/.gemini/antigravity/mcp_config.json` (macOS) / `%USERPROFILE%\.gemini\antigravity\mcp_config.json` (Windows)
+- Dùng `npx -p flowseeker flowseeker-mcp --workspace .` hoặc local build `node dist/cli/mcpServer.js --workspace .` để add FlowSeeker MCP server
+- Native Chat hiện có thể chèn `@mcp:flowseeker:` nhưng không hiện gợi ý tool/prompt tiếp theo; dùng natural language hoặc gọi tool name trực tiếp
+
+### Theia / OpenVSX-compatible
+Thường không có Native Chat và Language Model API.
+- Sidebar chat và direct provider là primary workflow
+- Copy Agent Prompt rất hữu ích trên các host này
+
+---
+
+## Graceful Degradation
+
+Khi host thiếu một API, FlowSeeker tự động fallback:
+
+1. **Không có Native Chat** → Sidebar chat làm primary interface. Copy Agent Prompt để paste context vào tool AI bất kỳ. Hoặc dùng **MCP Server** để integrate FlowSeeker vào native chat của IDE.
+2. **Không có Language Model API** → Dùng direct provider key đã cấu hình. Token counting dùng local BPE (cl100k_base).
+3. **Không có MCP APIs** → MCP-related UI elements bị ẩn hoặc hiển thị thông báo "not available on this host".
+4. **Không có Secrets Storage** → API key nhập per-session trong provider configuration.
+
+---
+
+## Update Notices
+
+FlowSeeker 0.1.7 and newer checks the npm registry at most once per day from the extension host. If a newer version is available, it shows a non-modal notice with actions to view npm or dismiss that version. Offline/proxy failures are silent.
+
+The standalone MCP server does not perform passive network checks during normal stdio startup. Use explicit CLI checks instead:
+
+```bash
+flowseeker-mcp --version
+flowseeker-mcp --check-updates
+```
+
+---
+
+## MCP Server (Cross-IDE Integration)
+
+FlowSeeker cung cấp một **standalone MCP server** cho phép tích hợp retrieval pipeline vào native chat của **bất kỳ IDE nào** hỗ trợ MCP: Cursor, Antigravity, Windsurf, Claude Desktop, v.v.
+
+### Khởi động MCP Server
+
+```bash
+# Chạy trực tiếp
+node dist/cli/mcpServer.js --workspace /path/to/repo
+
+# Hoặc qua npx (sau khi publish) — package tên `flowseeker`, bin là `flowseeker-mcp`
+npx -p flowseeker flowseeker-mcp --workspace /path/to/repo
+```
+
+### Cấu hình trong IDE
+
+#### Cursor
+
+Global config:
+
+- macOS: `~/.cursor/mcp.json`
+- Windows: `%USERPROFILE%\.cursor\mcp.json`
+
+Project config:
+
+- `.cursor/mcp.json` trong project root
+
+```json
+{
+  "mcpServers": {
+    "flowseeker": {
+      "command": "npx",
+      "args": ["-p", "flowseeker", "flowseeker-mcp", "--workspace", "."]
+    }
+  }
+}
+```
+
+#### Antigravity
+
+Config file:
+
+- macOS: `~/.gemini/antigravity/mcp_config.json`
+- Windows: `%USERPROFILE%\.gemini\antigravity\mcp_config.json`
+
+```json
+{
+  "mcpServers": {
+    "flowseeker": {
+      "command": "npx",
+      "args": ["-p", "flowseeker", "flowseeker-mcp", "--workspace", "."]
+    }
+  }
+}
+```
+
+If testing before npm publish, point Antigravity to the local build:
+
+```json
+{
+  "mcpServers": {
+    "flowseeker": {
+      "command": "node",
+      "args": [
+        "D:\\Developing\\FlowSeeker\\dist\\cli\\mcpServer.js",
+        "--workspace",
+        "D:\\Developing\\FlowSeeker"
+      ]
+    }
+  }
+}
+```
+
+#### Windsurf
+
+Config file:
+
+- macOS: `~/.codeium/windsurf/mcp_config.json`
+- Windows: `%USERPROFILE%\.codeium\windsurf\mcp_config.json`
+
+```json
+{
+  "mcpServers": {
+    "flowseeker": {
+      "command": "npx",
+      "args": ["-p", "flowseeker", "flowseeker-mcp", "--workspace", "."]
+    }
+  }
+}
+```
+
+#### VS Code / GitHub Copilot
+
+VS Code uses `.vscode/mcp.json` and root key `servers`:
+
+```json
+{
+  "servers": {
+    "flowseeker": {
+      "command": "npx",
+      "args": ["-p", "flowseeker", "flowseeker-mcp", "--workspace", "."]
+    }
+  }
+}
+```
+
+> Cursor, Antigravity, Windsurf, Claude Desktop, and Claude Code use `mcpServers`. VS Code/Copilot uses `servers`.
+
+#### Claude Desktop
+
+Thêm vào `claude_desktop_config.json`:
+
+```json
+{
+  "mcpServers": {
+    "flowseeker": {
+      "command": "npx",
+      "args": ["-p", "flowseeker", "flowseeker-mcp", "--workspace", "/path/to/repo"]
+    }
+  }
+}
+```
+
+> **Dev / local build:** Nếu đã clone repo và build từ source, dùng:
+>
+> ```json
+> {
+>   "mcpServers": {
+>     "flowseeker": {
+>       "command": "node",
+>       "args": ["/absolute/path/to/flowseeker/dist/cli/mcpServer.js", "--workspace", "/path/to/repo"]
+>     }
+>   }
+> }
+> ```
+
+### Available MCP Tools and Prompts
+
+| Tool | Description |
+|---|---|
+| `flowseeker_retrieve` | Run full retrieval pipeline, return Solve Packet with ranked evidence, coverage, token savings |
+| `flowseeker_guide` | Retrieve + return agent solve prompt for guidance |
+| `flowseeker_files` | Retrieve + return only top ranked file list (lightweight) |
+
+FlowSeeker also exposes MCP prompts for hosts that surface prompt suggestions:
+
+| Prompt | Maps to |
+|---|---|
+| `retrieve` | `flowseeker_retrieve` |
+| `guide` | `flowseeker_guide` |
+| `files` | `flowseeker_files` |
+
+Autocomplete UI is host-controlled. Antigravity currently may insert `@mcp:flowseeker:` and stop there; use natural language or direct tool names when no second-level suggestions appear.
+
+### Cách hoạt động
+
+1. IDE gửi tool call tới FlowSeeker MCP server
+2. FlowSeeker chạy headless pipeline (scan workspace → rank evidence → build solve packet)
+3. Kết quả trả về cho IDE's AI model tự động
+4. AI model reason over context và trả lời user trong native chat
+
+Ưu điểm: **dùng luôn model có sẵn của IDE** (Cursor dùng Claude/GPT tích hợp, Antigravity dùng model đã chọn). Không cần thêm API key riêng cho FlowSeeker.

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 FlowSeeker
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.