@promptweave/core
v0.0.1
Published
The Git for Prompts — Open-source AI Prompt Engineering Platform
Maintainers
Readme
PromptWeave
The Git for Prompts — Manage, test, and deploy AI prompts across multiple models.
PromptWeave is an open-source AI Prompt Engineering Platform that helps development teams manage the full lifecycle of AI prompts: authoring, versioning, testing, evaluation, and deployment.
Why PromptWeave?
As LLMs multiply (GPT, Claude, Gemini, DeepSeek, Qwen...), teams face a new challenge: managing hundreds of prompts across multiple models, environments, and experiments.
PromptWeave solves this with:
- Version Control — local/test/prod three-stage versioning (like Git branches for prompts)
- MetaPrompt — Define once, generate N variants for different models/platforms/experiments
- Multi-Environment Sync — Pull/Push with pluggable adapters to sync with any business system
- Playground — Test prompts against multiple models side-by-side (v1.1)
- Dataset & Evaluation — Batch testing with scoring and regression detection (v1.2)
- Prompt Workflow — DAG-based prompt chain orchestration (v2.0)
Quick Start
docker compose upComing soon. PromptWeave is under active development.
Tech Stack
| Layer | Technology | |-------|------------| | Backend | NestJS 11 + Fastify + TypeScript | | ORM | Prisma (PostgreSQL) | | Validation | Zod v3 + nestjs-zod | | Frontend | React 19 + Ant Design 6 | | Testing | Vitest | | Package Manager | pnpm |
Roadmap
| Version | Features | Status | |---------|----------|--------| | v1.0 | Prompt/Snippet/Provider CRUD + Versioning + Sync + MetaPrompt | In Progress | | v1.1 | Playground + Dataset + LCA Context Engine | Planned | | v1.2 | Batch Evaluation + A/B Comparison + Regression Testing | Planned | | v2.0 | Prompt Workflow (DAG orchestration, inspired by n8n) | Planned |
Unique Features
MetaPrompt Batch Variants
Define a base prompt and generate variants for every model/platform/experiment combination with one click. No competitor offers this.
LCA Smart Context (v1.1)
Powered by llm-context-attention — topic-aware context routing that saves ~50% tokens in multi-turn Playground conversations.
SyncAdapter Pattern
Pluggable adapters to sync prompt configurations with any business system via Pull/Push. Not locked into any specific SDK.
Contributing
Contributions are welcome! See CONTRIBUTING.md for details.
