local-first-llm
by joelnishanth
Coding Agents & IDEsBrief
Routes LLM requests to a local model (Ollama, LM Studio, llamafile) before falling back to cloud APIs.
Description
What This Skill Does
Routes LLM requests to a local model (Ollama, LM Studio, or llamafile) before falling back to cloud APIs. Routing decisions factor in data sensitivity, prompt complexity, and local provider availability. Token usage and cost savings are logged to a persistent dashboard.
Combines automatic routing logic with cost tracking, so users get privacy and savings benefits without manually deciding which model to use each time.
When to Use It
- Keeping sensitive prompts off cloud APIs
- Reducing OpenAI API costs on repetitive tasks
- Summarizing meeting transcripts with a local model
- Checking how much money local routing has saved this month
- Automatically routing complex queries to cloud while simple ones stay local
Example Workflow
User asks: summarize this meeting transcript and keep it private
- 1
Run check_local.py to detect available local providers
- 2
Run route_request.py with the prompt and token estimate; sensitivity check routes to local
- 3
Send the prompt to Ollama via its local API endpoint
- 4
Run track_savings.py log to record the token count and routing decision
- 5
Run dashboard.py to display updated cumulative savings
Meeting transcript summarized locally; dashboard shows 800 tokens saved and cost avoidance logged
Requirements
- Ollama, LM Studio, or llamafile installed and running locally
Installation
clawhub install joelnishanth/local-first-llmRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.