@pokutuna/pi-google-genai

A pi extension that adds Google GenAI grounding tools, built on the official @google/genai SDK.

Tools

| Tool | What it does | | ---------------- | ---------------------------------------------------------------- | | google_search | Search-grounded question answering via Google Search | | google_maps | Maps-grounded answers (places, nearby, routes, local businesses) | | url_context | Ask Gemini a question using specific http/https URLs as context | | deep_research | Agentic multi-step research via Gemini Deep Research (Vertex AI only) |

Grounding tools return the model's synthesized answer plus a Sources: section. Long outputs are truncated; the full raw response is saved to a session-scoped temp file.

google_search accepts an optional searchTypes array ("web_search" and/or "image_search"). Some models do not support image search grounding.

deep_research starts Gemini's Deep Research agent and returns an interactionId; it typically takes several minutes to finish. Use the ID without query to check its status. It is available only with Vertex AI. Its answer keeps the agent's own [cite: N] markers and ends with a Sources: list of every page it cited, as [N] title — url under those same numbers — so [cite: 6] in the text points at [6] in the list.

Default model: gemini-3.6-flash; Vertex AI location defaults to global.

Authentication

Gemini Developer API (API key)

Use either of these environment variables:

export GEMINI_API_KEY=your-key      # or GOOGLE_API_KEY

Alternatively, set apiKey in the config file. Pi's /login google credential is used when lookupPiConfig is true.

Vertex AI (Application Default Credentials)

Vertex AI authentication uses Application Default Credentials (ADC). For local development, authenticate with:

gcloud auth application-default login
export GOOGLE_CLOUD_PROJECT=my-gcp-project
# optional, defaults to "global"
export GOOGLE_CLOUD_LOCATION=us-central1

See the Application Default Credentials documentation for other setup options. This is also the required backend for deep_research.

Set GOOGLE_GENAI_USE_VERTEXAI=true or "auth": "vertex-ai" to force Vertex AI. Otherwise, an API key takes precedence when both are available.

Configuration

Optional JSON file: ~/.pi/agent/google-genai.json (or $PI_CODING_AGENT_DIR/google-genai.json if set).

{
  "auth": "vertex-ai",
  "project": "my-gcp-project",
  "model": "gemini-3.6-flash",
  "lookupPiConfig": false
}

Config values override environment variables. Unknown or invalid fields produce warnings. apiKey must be a literal value; $ENV_VAR and command interpolation are not supported.

To enable lookup of pi's Google configuration, use:

{
  "lookupPiConfig": true
}

Pi configuration lookup

lookupPiConfig defaults to false; the extension then uses only its own settings and environment variables. Set it to true to use pi's current Google provider and credentials:

• current provider google-vertex uses Vertex AI;
• pi's google / google-vertex credentials, including GOOGLE_CLOUD_API_KEY, are used when available;
• otherwise Gemini API authentication is preferred over Vertex AI authentication (so a current google provider resolves to its API key too).

Explicit settings in google-genai.json still take precedence.

Command

• /google-genai or /google-genai status — show the resolved configuration (never the key itself) and any warnings.
• /google-genai help — usage and configuration instructions.