@fradser/pi-context

Retrieve code context for any repo, library, or natural-language query via DeepWiki, Context7, Exa, git clone, and web search+fetch. Native pi tools (context_deepwiki/context_context7/context_exa) call the public REST APIs directly — no MCP servers.

Packages

Package details

extension

Install @fradser/pi-context from npm and Pi will load the resources declared by the package manifest.

$ pi install npm:@fradser/pi-context
Package
@fradser/pi-context
Version
0.5.1
Published
Aug 19, 2026
Downloads
327/mo · 327/wk
Author
fradser
License
MIT
Types
extension
Size
26.3 KB
Dependencies
0 dependencies · 3 peers
Pi manifest JSON
{
  "extensions": [
    "./index.ts"
  ]
}

Security note

Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.

README

Context Pi Package

Version: 0.5.0

Retrieve code context for any repo, library, or natural-language query via 5 methods: DeepWiki, Context7, Exa, git clone, and web search+fetch.

The package does not ship a skill. Native tool guidance is injected into the system prompt, and the /context command loads the full workflow into the current turn.

Installation

pi install npm:@fradser/pi-context
# or from this repository:
pi install /path/to/pi-packages/packages/context

Usage

Run /context followed by a natural-language question, repository slug, library name, or several targets. For example:

/context React 19 server actions error handling
/context facebook/react
/context facebook/react zustand --method=deepwiki,context7

--method= accepts deepwiki,context7,exa,clone,web,all (comma-separated). With no target, the command reads dependency manifests in the current directory (package.json, pyproject.toml, go.mod, Cargo.toml) and uses detected dependencies as targets.

The full workflow (target classification, per-method process, fallbacks, selection guide) lives in references/workflow.md and is injected into the turn when /context runs. Lightweight tool-selection guidance is always present in the system prompt.

Runtime requirements

Method Availability Notes
DeepWiki Always (native tool context_deepwiki) No API key required
Context7 Always (native tool context_context7) Optional CONTEXT7_API_KEY (Bearer) for per-key quota
Exa Native tool context_exa Requires EXA_API_KEY env var
Git clone Always Pi bash + read
Web fetch Always curl/HTTP via bash

All three retrieval tools are registered by this package's root extension entry (index.ts) and call the public REST APIs directly — no MCP servers needed.

Methods (summary)

  1. DeepWiki — AI-generated repo docs via context_deepwiki (mode: structure | contents | ask).
  2. Context7 — up-to-date library docs via context_context7 (query + optional topic).
  3. Exa — web/code search via context_exa (query + optional numResults, requires EXA_API_KEY).
  4. Git clonegit clone --depth=1 to /tmp, read key files, clean up.
  5. Web search+fetchbash + curl against official docs, GitHub issues, changelogs.

If a native tool is unavailable or errors, fall back to clone/web. The /context workflow reference contains the full selection matrix and per-method process.

Optional manual brief

agents/context-researcher.md is an optional manual prompt brief for an isolated research pass. Pi does not load it as an invocable agent; use the /context command for the executable workflow.

Structure

context/
├── index.ts                     # Package-root extension entry point
├── extensions/
│   ├── context-tools.ts         # Native pi tools: context_deepwiki / context_context7 / context_exa
│   └── context-command.ts       # /context command + system-prompt guidance
├── references/
│   └── workflow.md              # Full workflow injected by /context
├── agents/
│   └── context-researcher.md    # Isolated research brief (manual)
└── README.md

Prerequisites

  • Pi with extensions + custom tools
  • Network access for external lookups
  • EXA_API_KEY for the Exa method (optional for the other four)

Best Practices

  • Local first: check local context (package.json, imports) before external lookups.
  • Combine methods: DeepWiki / clone for architecture, Context7 for API surface, Exa / web for community patterns.
  • Verify sources: cross-reference fetched content against actual code.
  • Clean up: remove /tmp clones when finished.

License

MIT

Author

Frad LEE (fradser@gmail.com)