Venice Image is a modular tool suite for intelligent agents that interfaces with the Venice AI API. It provides four primary capabilities:

1. Image Generation: Prompt-based creation using models like Fluently XL, Flux Dev, Lustify SDXL, Pony Realism, and Venice SD35.
2. Image Enhancement: Stylizing or refining existing images without changing resolution.
3. Image Upscaling: Increasing resolution by 2x or 4x.
4. Image Vision: Generating detailed textual descriptions of images for content understanding or indexing.

All tools handle authentication, rate limiting, and error management automatically. Generated or processed images are stored in S3-compatible storage, and the suite returns stable URLs for consumption.

IntentKit is an intent-based AI agent platform designed for blockchain and cryptocurrency operations. Key features include:

• Intent-based Architecture: Agents understand and execute specific user intents.
• Blockchain & Crypto: Native support for multiple networks and built-in tools for DeFi, trading, and token operations.
• Extensible Tools: A modular system with 30+ pre-built tools covering DeFi (Uniswap, 1inch, Enso, LiFi), Data (DexScreener, CoinGecko, DefiLlama, CryptoCompare), Social (Twitter, Telegram, Slack), Blockchain (CDP, Moralis), and AI (OpenAI, Heurist, Venice AI).
• Multi-platform Support: Integrations for Telegram, Twitter, Slack, and standard APIs.
• AI Frameworks: Powered by LangChain and LangGraph.

Venice Image Generation provides prompt-based image creation using various AI models via the Venice AI API. Each model is accessible as a specific sub-tool within the suite, but they all share a unified interface and core set of options.

Supported Models:

• Fluently XL: Professional realism, lighting, and artistic work.
• Flux Dev: Artistic research and innovative workflows.
• Flux Dev Uncensored: Unrestricted/NSFW generation.
• Lustify SDXL: Photorealistic, NSFW/SFW versatility.
• Pony Realism: High-detail anime/character design (optimized for Danbooru tags).
• Venice SD35/Stable Diffusion 3.5: Artistic, illustrative, or design content.

IntentKit uses a modular architecture divided into four distinct layers to separate concerns between external interfaces and internal AI orchestration:

• Entrypoint Layer: Acts as the interface for external services (e.g., Twitter, Telegram). It uses adapters for I/O transformations, rate limiting, and input/output modifications.
• LangGraph Layer: The core orchestration engine. It manages LLM interactions, prompt engineering, tool execution flows, and maintains both thread-specific and broad agent memory.
• Processing Layer: Handles the runtime for Tools and Memory.
• Storage Layer: Manages persistent data, including agent configurations, secure credentials, agent state, and the memory store.

LiFi Tools enable cross-chain token transfers and swaps using the LiFi protocol. The suite includes tools for retrieving real-time quotes and executing transactions via a CDP wallet integration.

Available Tools

• token_quote: Get real-time quotes for token transfers and swaps. This tool is available for public access and does not require a wallet.
• lifi_token_execute: Execute token transfers and swaps. This tool requires a configured and funded CDP wallet.

Prerequisites

• A configured and funded CDP wallet.
• An agent with LiFi tools enabled.
• Sufficient token balance for the swap/transfer amount.
• Sufficient native gas tokens on the relevant networks for transaction fees.

XMTP Tools allow you to create XMTP protocol transaction requests (ETH or ERC20 transfers) on the Base mainnet. These requests are sent to users for signing.

Key Innovation: Response Format These tools use the content_and_artifact response format. Instead of returning a simple string, the tool returns a tuple: (content_message, List[ChatMessageAttachment]).

• content_message: The conversational text sent to the user.
• ChatMessageAttachment: A list containing the XMTP transaction data with type "xmtp".

To run IntentKit locally, you need Docker, Docker Compose, and Docker Buildx. On macOS, you can install these using Homebrew. Additionally, you must configure Docker to find plugins by updating your ~/.docker/config.json file.

brew update
brew install colima docker-compose docker-buildx

Update ~/.docker/config.json

{
  "cliPluginsExtraDirs": [
    "/opt/homebrew/lib/docker/cli-plugins"
  ]
}

Start the runtime

colima start

To enable Langfuse for agent observability, set the following environment variables. Langfuse is automatically enabled as soon as both LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY are present. Each conversation is automatically grouped into a single session in the Langfuse UI.

1. Sign up at cloud.langfuse.com (or use your self-hosted instance).
2. Create a project and generate a key pair (public and secret keys) in Settings → API Keys.
3. Set the environment variables listed below. If self-hosting, you must also set LANGFUSE_BASE_URL (or the legacy LANGFUSE_HOST).

LANGFUSE_PUBLIC_KEY=pk-lf-...
LANGFUSE_SECRET_KEY=sk-lf-...
# LANGFUSE_BASE_URL=https://cloud.langfuse.com   # optional, defaults to Langfuse Cloud