Letta v0.5 release
🧰 Dynamic model listing and multiple providers
In Letta v0.5, model providers (e.g. OpenAI, Ollama, vLLM, etc.) are now enabled using environment variables, where multiple providers can be enabled at a time. When a provider is enabled, all supported LLM and embedding models will be listed as options to be selected in the CLI and ADE in a dropdown.
For example for OpenAI, you can simply get started with:
> export OPENAI_API_KEY=...
> letta run
? Select LLM model: (Use arrow keys)
» letta-free [type=openai] [ip=https://inference.memgpt.ai]
gpt-4o-mini-2024-07-18 [type=openai] [ip=https://api.openai.com/v1]
gpt-4o-mini [type=openai] [ip=https://api.openai.com/v1]
gpt-4o-2024-08-06 [type=openai] [ip=https://api.openai.com/v1]
gpt-4o-2024-05-13 [type=openai] [ip=https://api.openai.com/v1]
gpt-4o [type=openai] [ip=https://api.openai.com/v1]
gpt-4-turbo-preview [type=openai] [ip=https://api.openai.com/v1]
gpt-4-turbo-2024-04-09 [type=openai] [ip=https://api.openai.com/v1]
gpt-4-turbo [type=openai] [ip=https://api.openai.com/v1]
gpt-4-1106-preview [type=openai] [ip=https://api.openai.com/v1]
gpt-4-0613 [type=openai] [ip=https://api.openai.com/v1]
...
Similarly, if you are using the ADE with letta server, you can select the model to use from the model dropdown.
# include models from OpenAI
> export OPENAI_API_KEY=...
# include models from Anthropic
> export ANTHROPIC_API_KEY=...
# include models served by Ollama
> export OLLAMA_BASE_URL=...
> letta server
We are deprecating the letta configure and letta quickstart commands, and the the use of ~/.letta/config for specifying the default LLMConfig and EmbeddingConfig, as it prevents a single letta server from being able to run agents with different model configurations concurrently, or to change the model configuration of an agent without re-starting the server. This workflow also required users to specify the model name, provider, and context window size manually via letta configure.
🧠 Integration testing for model providers
We added integration tests (including testing of MemGPT memory management tool-use) for the following model providers, and fixed many bugs in the process:
📊 Database migrations
We now support automated database migrations via alembic, implemented in #1867. You can expect future release to support automated migrations even if there are schema changes.
Read the full v0.5 changelog on GitHub.
Traditional LLMs operate in a stateless paradigm—each interaction exists in isolation, with no knowledge carried forward from previous conversations. Agent memory solves this problem.
As AI agents become more sophisticated, understanding how to design and manage their context windows (via context engineering) has become crucial for developers.
Although RAG provides a way to connect LLMs and agents to more data than what can fit into context, traditional RAG is insufficient for building agent memory.
Today we're announcing Letta Filesystem, which provides an interface for agents to organize and reference content from documents like PDFs, transcripts, documentation, and more.
We've releasing new client SDKs (support for TypeScript and Python) and upgraded developer documentation
Introducing Agent File (.af): An open file format for serializing stateful agents with persistent memory and behavior.
Introducing the Letta Agent Development Environment (ADE): Agents as Context + Tools
Letta v0.6.4 adds Python 3.13 support and an official TypeScript SDK.
Letta v0.5.2 adds tool rules, which allows you to constrain the behavior of your Letta agents similar to graphs.
We're excited to announce the Letta Leaderboard, a comprehensive benchmark suite that evaluates how effectively LLMs manage agentic memory.
Memory blocks offer an elegant abstraction for context window management. By structuring the context into discrete, functional units, we can give LLM agents more consistent, usable memory.
