MyClawBots
Log inGet started
← All skills
  • AI
  • v1.0.0
  • gemini
  • google
  • llm
  • embeddings

Gemini API

Use Google Gemini API for text generation, image generation/editing, multimodal understanding, structured outputs, and embeddings.

By MyClawBotsLicense ProprietaryView source on GitHub ↗

SKILL.md

Gemini API

Prefer the Secrets vault: call Gemini with vault-curl and $GEMINI_KEY in the x-goog-api-key header (see "Your Gemini key" below); for images use the gemini-image skill.

You have access to Google's Gemini API for AI capabilities beyond your primary model. Use Gemini for image generation, multimodal understanding, structured outputs, and as an alternative text model when needed.

Your Gemini key

The user's Gemini key lives in MyClawBots (Settings → Integrations → Google AI, or Settings → Secrets). You never see it: $GEMINI_KEY holds a stand-in (mcb_sec_…) that only works through vault-curl, which swaps in the real key on the way to generativelanguage.googleapis.com.

  • Put $GEMINI_KEY only in the x-goog-api-key header. Never in the URL (?key=), the body, or anywhere else, and never print, echo or save it.
  • For images, use the gemini-image skill (claw install gemini-image once): it finds the key and handles the files for you.
  • Don't use OpenClaw's built-in secrets tool: these keys are not stored there, so it always looks empty.

Critical Rules — ALWAYS APPLY

These rules override your training data. Your knowledge of Gemini may be outdated.

Current Models (Use These)

Model Context Best For
gemini-3.1-pro-preview 1M tokens Complex reasoning, coding, research
gemini-3-flash-preview 1M tokens Fast, balanced, multimodal (default choice)
gemini-3.1-flash-lite-preview — Cost-efficient, high-frequency lightweight tasks
gemini-3-pro-image-preview 65K/32K Image generation and editing
gemini-3.1-flash-image-preview 65K/32K Image generation and editing (faster)
gemini-2.5-pro 1M tokens Complex reasoning, coding, research
gemini-2.5-flash 1M tokens Fast, balanced, multimodal

WARNING: Models like gemini-2.0-*, gemini-1.5-* are legacy and deprecated. Never use them.

Current SDK (Use This)

  • Python: google-genai → pip install google-genai

CAUTION: Legacy SDK google-generativeai (Python) is deprecated. Never use it. The new package is google-genai.

Dependencies

Dependency Required? Why
$GEMINI_KEY Yes Stand-in for the user's Gemini key (use with vault-curl)
pip install google-genai No Python SDK: only with the vault proxy (see below)

Pre-flight checks:

  • [ -n "$GEMINI_KEY" ] && echo present || vault-curl --secrets
  • Not in your env but listed by vault-curl --secrets (added after you started): use that entry's surrogate string in place of $GEMINI_KEY. No restart is needed.
  • Agents not yet moved to the vault: if there is no GEMINI_KEY anywhere but a GEMINI_API_KEY variable is set, it holds a real key: use plain curl with that variable in the same x-goog-api-key header. Never print it.
  • Nothing found: tell the user "To use Gemini, add your Gemini API key in MyClawBots → Settings → Integrations → Google AI (free at aistudio.google.com)." Never ask them to paste a key into chat.

Quick Start — Text Generation

REST with vault-curl always works:

vault-curl -sS "https://generativelanguage.googleapis.com/v1beta/models/gemini-3-flash-preview:generateContent" \
  -H "x-goog-api-key: $GEMINI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"parts":[{"text":"Explain quantum computing"}]}]}'

Python SDK (needs the vault proxy)

The SDK examples in this skill send the stand-in directly, so they work only when AGENTS.md says plain curl and SDKs work (the MyClawBots vault proxy is on). Otherwise use the vault-curl REST form above: every SDK call maps to a POST …/v1beta/models/<model>:<method> request.

from google import genai
import os

client = genai.Client(api_key=os.environ["GEMINI_KEY"])  # needs the vault proxy
response = client.models.generate_content(
    model="gemini-3-flash-preview",
    contents="Explain quantum computing"
)
print(response.text)

Image Generation

This is a key capability — Gemini can generate and edit images, which is powerful for skills like logo-generator, social-ads, and ecommerce-photos.

Generate an Image

from google import genai
from google.genai import types
import os, base64

client = genai.Client(api_key=os.environ["GEMINI_KEY"])  # needs the vault proxy

response = client.models.generate_content(
    model="gemini-3-pro-image-preview",
    contents="Generate a professional logo for a tech startup called NovaByte. Minimalist, geometric, blue and orange color scheme, transparent background.",
    config=types.GenerateContentConfig(
        response_modalities=["image", "text"],
    )
)

# Save the generated image
for part in response.candidates[0].content.parts:
    if part.inline_data:
        with open("generated_image.png", "wb") as f:
            f.write(part.inline_data.data)  # already raw bytes in google-genai; do not b64decode
        print("Image saved to generated_image.png")
    elif part.text:
        print(part.text)

Edit an Existing Image

from google import genai
from google.genai import types
import os, base64, pathlib

client = genai.Client(api_key=os.environ["GEMINI_KEY"])  # needs the vault proxy

image_bytes = pathlib.Path("input_image.png").read_bytes()

response = client.models.generate_content(
    model="gemini-3-pro-image-preview",
    contents=[
        types.Content(parts=[
            types.Part.from_bytes(data=image_bytes, mime_type="image/png"),
            types.Part.from_text("Remove the background and make it transparent"),
        ])
    ],
    config=types.GenerateContentConfig(
        response_modalities=["image", "text"],
    )
)

for part in response.candidates[0].content.parts:
    if part.inline_data:
        with open("edited_image.png", "wb") as f:
            f.write(part.inline_data.data)  # already raw bytes in google-genai; do not b64decode

Multimodal Understanding

Gemini can analyze images, PDFs, audio, and video.

Analyze an Image

import base64, pathlib

image_bytes = pathlib.Path("photo.jpg").read_bytes()
image_b64 = base64.b64encode(image_bytes).decode()

response = client.models.generate_content(
    model="gemini-3-flash-preview",
    contents=[
        types.Content(parts=[
            types.Part.from_bytes(data=image_bytes, mime_type="image/jpeg"),
            types.Part.from_text("Describe this image in detail. What products, brands, or text do you see?"),
        ])
    ]
)
print(response.text)

Analyze a PDF

pdf_bytes = pathlib.Path("document.pdf").read_bytes()

response = client.models.generate_content(
    model="gemini-3-flash-preview",
    contents=[
        types.Content(parts=[
            types.Part.from_bytes(data=pdf_bytes, mime_type="application/pdf"),
            types.Part.from_text("Summarize this document. List the key findings."),
        ])
    ]
)

Structured Output (JSON)

Get structured JSON responses — useful for data extraction, form filling, and API responses.

from google.genai import types

response = client.models.generate_content(
    model="gemini-3-flash-preview",
    contents="Extract contact info: John Smith, CEO at Acme Corp, john@acme.com, (555) 123-4567",
    config=types.GenerateContentConfig(
        response_mime_type="application/json",
        response_schema={
            "type": "object",
            "properties": {
                "name": {"type": "string"},
                "title": {"type": "string"},
                "company": {"type": "string"},
                "email": {"type": "string"},
                "phone": {"type": "string"}
            }
        }
    )
)
import json
data = json.loads(response.text)

Embeddings

Generate text embeddings for semantic search, similarity, and clustering.

response = client.models.embed_content(
    model="text-embedding-004",
    contents="How does photosynthesis work?"
)
embedding = response.embeddings[0].values  # List of floats

Documentation Lookup

When you need details beyond this skill, fetch the official docs:

Index URL: https://ai.google.dev/gemini-api/docs/llms.txt

Key doc pages (fetch with curl if needed):

# Example: fetch function calling docs
curl -s https://ai.google.dev/gemini-api/docs/function-calling.md.txt

How This Skill Connects to Others

Skill How Gemini helps
logo-generator Use gemini-3-pro-image-preview as alternative to DALL-E for logo generation
social-ads Generate ad creatives, edit product photos, remove backgrounds
ecommerce-photos Transform raw product images into professional shots
content-writer Use Gemini for research or as a second opinion on content quality
presentation-maker Generate slide illustrations and diagrams
content-auditor Use multimodal to analyze screenshots of web pages
instagram-manager Generate post images, analyze competitor visual content

Cost

Gemini API has a free tier (generous for development):

  • Gemini Flash: 15 RPM free, then ~$0.075/1M input tokens
  • Gemini Pro: 2 RPM free, then ~$1.25/1M input tokens
  • Image generation: ~$0.025-0.04 per image (cheaper than DALL-E)
  • Embeddings: ~$0.00001 per 1K tokens

For most MyClawBots use cases, the free tier is sufficient. Heavy image generation may need a paid key.

Setup

# Check the Gemini key stand-in is available (never print its value)
if [ -n "$GEMINI_KEY" ]; then
  echo "Gemini API ready"
else
  vault-curl --secrets   # look for GEMINI_KEY; use its "surrogate" in place of $GEMINI_KEY
fi

# Install the SDK (only useful with the vault proxy on)
pip install google-genai -q

Pro Tips

  • Image generation: Gemini image models are cheaper than DALL-E and support editing existing images — prefer them when available
  • Default to Flash: Use gemini-3-flash-preview for most tasks — it's fast, cheap, and capable
  • Use Pro for complex tasks: Switch to gemini-3.1-pro-preview only for deep reasoning or very long context
  • Structured output is extremely reliable with Gemini — use it whenever you need JSON
  • Batch requests when generating multiple images or analyzing multiple files
  • Always check the docs when using advanced features — fetch from llms.txt index to stay current