Report Abuse

Basic Information

Potpie is an open-source Prompt-To-Agent platform that creates AI agents specialized for a specific codebase to automate code analysis, testing, debugging and development tasks. The repository implements a system that builds a knowledge graph of code relationships so agents can reason about functions, files and architecture. It provides a UI submodule, a FastAPI backend, Celery workers and Docker compose orchestration and requires Python 3.10, PostgreSQL, Neo4j and Redis. The project exposes REST endpoints to parse repositories, monitor parsing status, list and run available agents, create conversations and send messages to agents. It also supports provider model configuration via environment variables and integration points for custom tools, enabling both local development workflows and production deployment.

Links

Categorization

App Details

Features
Prebuilt agents include debugging, codebase Q&A, code changes analysis, integration test generation, unit test generation, low level design and code generation agents. A tooling layer offers functions such as get_code_from_probable_node_name, get_code_from_node_id, get_code_from_multiple_node_ids, ask_knowledge_graph_queries, get_nodes_from_tags, get_code_graph_from_node_id/name, change_detection and get_code_file_structure to let agents query the knowledge graph and retrieve code context. Integrations include a VSCode extension and Slack app for in-editor and team collaboration. Installation and runtime are driven by scripts (start.sh, stop.sh, start.ps1, stop.ps1), a potpie-ui submodule with pnpm build/start steps, and environment configuration for inference and chat models, provider API keys and database connections.
Use Cases
Potpie helps developer teams speed onboarding, debugging and code review by providing agents that understand a repository"s structure and context. It can generate test plans and test code, propose low level designs for new features, evaluate blast radius of code changes and provide stepwise debugging guidance grounded in the codebase. Teams can integrate agents into CI/CD and automation flows via API keys and REST APIs to parse repos, create project ids and conversations, and programmatically interact with agents. The VSCode and Slack integrations let developers ask questions and get contextual suggestions where they work. Custom agents can be authored via a single prompt and extended by adding tools in the app/modules/intelligence/tools directory.

Please fill the required fields*