Comparisons¶
- OKF Generator vs OpenWiki — code-to-knowledge vs clipboard-to-wiki
- OKF Generator vs okft — bundle producer vs bundle consumer/validator
OKF Generator vs OpenWiki — Comparison¶
Quick Summary¶
| Dimension | OKF Generator | OpenWiki |
|---|---|---|
| Purpose | Codebase → structured knowledge graph for AI agents | Clipboard → personal wiki for humans |
| Input | Source code (13 langs, 20+ manifests) | Clipboard (text, URLs, images) |
| Output | Portable markdown bundles, CLI consumable | SQLite DB, desktop GUI |
| AI Dependency | Zero-AI by design (pure AST parsing) | AI-critical (LLM required for every op) |
| Delivery | PyPI package, CLI, MCP server | Tauri desktop app (DMG/MSI) |
| Graph | Typed edges: calls, extends, depends-on | TF-IDF cosine similarity on tag vectors |
| Target User | Developers, AI agents, CI/CD | Information workers, content collectors |
| Language | 10 programming languages | Natural language via LLM |
| Offline | Fully offline | Needs API keys |
| Stars | ~0 (new) | 513 |
| Platform | Cross-platform (Python) | macOS + Windows (Tauri) |
| License | MIT | MIT |
Feature Comparison¶
| Feature | OKF Generator | OpenWiki |
|---|---|---|
| Code parsing (AST) | ✅ 13 languages, tree-sitter | ❌ |
| Dependency/manifest parsing | ✅ 20+ formats | ❌ |
| Call graph extraction | ✅ Cross-file resolution | ❌ |
| Cross-reference linking | ✅ imports→deps + calls | ❌ |
| Portable bundle output | ✅ Flat markdown files | ❌ (SQLite) |
| Bundle diff/versioning | ✅ okf diff |
❌ |
| Training data generation | ✅ JSONL pairs (5 types) | ❌ |
| MCP protocol | ✅ Stdio + HTTP/SSE | ✅ SQLite via npx |
| AI enrichment (optional) | ✅ 4 modes, multi-provider | ❌ (not optional) |
| Clipboard monitoring | ❌ | ✅ Background daemon |
| URL content extraction | ❌ | ✅ WeChat, X, YouTube, etc. |
| OCR / image capture | ❌ | ✅ macOS native + WinRT |
| Auto wiki compilation | ❌ | ✅ AI-driven |
| Insight reports | ❌ | ✅ Weekly, 7-dimension |
| Knowledge graph viz | ✅ D3.js (static HTML) | ✅ D3 force graph (in-app) |
| GUI | ❌ (CLI only) | ✅ Full desktop GUI |
| Agent integrations | ✅ 6 agents (skills) | ✅ MCP for Claude Desktop |
| Preference learning | ❌ | ✅ Like/dismiss feedback loop |
| CI/CD integration | ✅ GitHub Action, Docker | ❌ |
Pros & Cons¶
OKF Generator¶
Pros - Zero AI dependency — fully offline, no API costs, deterministic - 13-language parser coverage with rich detail (signatures, params, inheritance, call graphs) - Portable markdown bundles — easy to diff, version, cache, grep - Designed for AI agent consumption — lookup, search, filter by type/tag/file - Generates JSONL training pairs for fine-tuning - MCP server + 6 agent integrations (Claude Code, OpenCode, Cursor, Copilot, Windsurf, Cline) - 20+ dependency manifest formats - CI/CD ready (GitHub Action, Docker, pre-commit)
Cons - CLI only — no GUI, higher onboarding friction - Narrow focus: source code only, no natural language content - No user feedback loop or preference learning - Smaller community / fewer stars - Tree-sitter dependency for non-Python langs (native compilation needed)
OpenWiki¶
Pros - Beautiful desktop GUI with popup capture, graph viz, calendar timeline - Broad input: any clipboard content (text, URLs, images, YouTube) - AI enrichment: summarization, tagging, insight reports, attention analysis - Privacy-first: local SQLite, no cloud - Knowledge linting (orphan detection, broken links) - Preference learning engine - 513 GitHub stars, active community - MCP protocol integration for Claude Desktop
Cons - Completely dependent on AI — dead without API key and internet - Proprietary SQLite format — not portable, diffable, or greppable - Cannot analyze source code structurally (treats it as plain text) - Non-deterministic — same input can produce different wiki page output - No CI/CD pipeline integration - No training data generation - macOS/Windows only — no Linux, no CLI mode - Platform-specific build issues (unsigned DMG, permission prompts)
Architecture Differences¶
OKF Generator:
Source Code → AST Parser → Concept Extraction → Linker (calls/deps)
→ OKF Bundle (markdown) → [Lookup | Diff | Pairs | Viz | MCP]
OpenWiki:
Clipboard → AI Summarize → AI Assess → AI Compile → Wiki Pages (SQLite)
→ Knowledge Graph (TF-IDF cosine) → [Search | Graph | Reports | MCP]
Impact & Synergy Analysis¶
They Do Not Compete¶
OKF Generator and OpenWiki solve fundamentally different problems for different users:
| Dimension | OKF Generator | OpenWiki |
|---|---|---|
| Persona | Developer, CI/CD, AI agent | Information worker, knowledge collector |
| Problem | "How do I give my AI agent codebase context?" | "How do I organize everything I read?" |
| Workflow | okf generate → okf lookup in terminal |
Copy → popup → wiki pages in GUI |
Potential Synergy¶
-
Code snippet capture in OpenWiki: OpenWiki treats code as plain text. Routing code content through OKF Generator's tree-sitter parsers before wiki compilation would produce structured wiki pages with signatures, params, and call graphs.
-
OKF bundle import into OpenWiki: OpenWiki users could import OKF bundles as structured project knowledge, queryable alongside their personal notes.
-
OpenWiki-style insight reports for codebases: OKF Generator could adopt OpenWiki's attention analysis and weekly report patterns for code health trends (e.g., "this week you deleted 500 lines of dead code").
-
Shared MCP ecosystem: Both projects implement MCP servers. A unified MCP tool that queries both personal wiki + code knowledge would give agents a holistic view.
When to Use Which¶
| If you need... | Use |
|---|---|
| Give Claude/OpenCode context about a codebase | OKF Generator |
| Capture and organize articles, tweets, notes | OpenWiki |
| Generate training data from source code | OKF Generator |
| Spot code smells and call-graph issues | OKF Generator |
| Track what you read and get AI insights | OpenWiki |
| CI/CD pipeline that diffs code knowledge | OKF Generator |
| Desktop app with GUI and popup capture | OpenWiki |
| Both — use them side-by-side, they don't overlap | Both |