LLM Framework is an open-source environment for local language models. It organizes work into separate workspaces so instructions, documents, code, memory, and conversations do not unintentionally mix.
Why the framework exists
A local model becomes more useful when it can do more than sit behind one general chat window. This project adds structure, local search, and visible boundaries around the model you choose through Ollama.
What is already working
Separated workspaces
Domains and sub-domains keep prompts, documents, memory, conversations, and permissions inside the selected context.
Local knowledge
Documents and read-only repository snapshots can be searched locally with references back to the retrieved source.
Measured settings
The model optimizer compares context settings and shows the tradeoffs before any selected change is applied.
Designed to run locally
Docker Compose starts the application, PostgreSQL with pgvector, and Ollama on the device. You choose the model and keep control of the working environment.
Current boundary: the project is under active development and is not yet hardened for multiple users or direct public-internet exposure.