symbolic-scribe
github.com/ruvnet/symbolic-scribe- Category
- Productivity
- Rank
- No. 1977Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- ruvnet
- GitHub
- 88 stars
- Date
About
rUv's tool for symbolic note-taking — captures structured knowledge with named relations between concepts for graph-based recall.
What it does
Symbolic Scribe turns loose prompts into structured candidates, checks ambiguity and safety risks, compares multiple objectives, and produces an integrity receipt. Its browser interface also provides mathematical templates, model previews, and an editable evaluation matrix for testing candidates through OpenRouter.
Why it's ranked here
The strongest case is its deterministic browser-side compiler combined with measured model evaluation. It refuses an optimization when accuracy, safety, or schema validity regresses. The scope is unusually broad, but some security language needs care because its client-bundled HMAC key cannot prove authenticity.
What's good
Core analysis runs locally without sending prompt text away. Candidate grading can test JSON shape, required fields, assertions, refusals, output length, latency, and cost across models. The compiler preserves numbers, entities, and constraints through a drift report, while the firewall returns explicit allow, log, approval, or block decisions.
Tradeoffs
Real model evaluation requires OpenRouter access and incurs external calls, latency, and possible cost. The integrity receipt is tamper-evident, not an authenticity signature, because its HMAC key ships to the browser. Mathematical templates may add ceremony for ordinary prompting, and the broad red-team focus assumes users can define responsible testing boundaries.
How to use it well
Use it for repeatable prompt engineering, AI safety experiments, and authorized red-team work where constraints, regression checks, and audit artifacts matter. Start from a template or rough prompt, inspect compiler findings, then validate frontier candidates against an editable test suite. It does not replace server-side signing, model hosting, or a full vulnerability-management workflow.
Technical notes+
package.json defines a Vite, React, and TypeScript application plus Jest scripts, Rust-to-WASM build and test commands, and a published harness invocation documented in README.md. src/App.tsx registers browser routes for generation, optimization, templates, documentation, settings, and about pages. src/pages/Optimizer.tsx initializes the WASM compiler, debounces live analysis, runs firewall checks, optimizes with a token budget and witness key, records outcomes, and optionally activates RuVector-backed recall and routing. src/services/promptEval.ts injects the model caller, grades candidate-model-test cells, replaces proxy metrics with measured results, and recomputes weighted composites. src/pages/Templates.tsx loads Markdown templates through Vite glob imports and applies firewall verdicts to red-team and blue-team entries.
Observed
- License
- MIT
- Primary languages
- TypeScript frontend with a Rust core compiled to WebAssembly
- Web interface
- Vite and React browser application with generator, compiler, templates, documentation, and settings routes
- Install surface
- Source installation uses npm; generated WebAssembly is committed so the web build does not require Rust
- CLI surface
- An npm-published repository agent harness exposes doctor, repository triage, and release-check commands
- Model interface
- Optional OpenRouter integration supports model discovery, prompt previews, and multi-model evaluation
- Testing surface
- Package scripts include Jest tests, coverage runs, native Rust unit tests, and WASM benchmarks
Read from README.md, package.json, src/App.tsx, src/main.tsx, src/vite-env.d.ts, src/lib/utils.ts, src/pages/Index.tsx, src/pages/About.tsx, src/hooks/use-toast.ts, src/pages/Settings.tsx, src/pages/Optimizer.tsx, src/pages/Templates.tsx, src/components/MainNav.tsx, src/services/promptEval.ts, src/pages/Documentation.tsx.
What it can do
Create symbolic notes with structured knowledge
User concepts and ideas → Structured symbolic notes
Define named relations between concepts
Two or more concepts → Named relationship connections
Build knowledge graphs from notes
Symbolic notes with relations → Visual knowledge graph
Search and recall information using graph structure
Query or concept name → Related concepts and connections
Navigate concept relationships
Starting concept → Connected concepts and relationship paths
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.