Product

AI Providers

Cloud providers, local Ollama, and LiteLLM/Bedrock setup.

TRCAA talks to whichever AI backend you choose, from a fully offline local model to any of the major cloud providers. Add or switch providers any time from Settings → AI Providers.

Supported providers

Provider	API URL	Notes
Ollama (local)	http://localhost:11434	No key needed — fully offline
OpenAI	api.openai.com/v1	Requires API key
Anthropic	api.anthropic.com	Requires API key
Google Gemini	generativelanguage.googleapis.com	Requires API key
Mistral	api.mistral.ai/v1	Requires API key
Azure OpenAI	<resource>.openai.azure.com	Requires API key
AWS Bedrock	via LiteLLM proxy	See below
Custom	any OpenAI-compatible URL	See below

Whichever provider you pick, PII detection and redaction runs first — see PII Detection. Nothing unredacted leaves your machine.

Offline (Ollama)

Fully local, zero network calls beyond localhost. Pull a model and point TRCAA at it — Settings → Ollama can also do hardware detection and pull models for you directly.

ollama pull llama3.2:3b   # good for most hardware (≥8 GB RAM)
ollama pull llama3.1:8b   # better quality (≥16 GB RAM)

Custom (any OpenAI-compatible endpoint)

Beyond the named providers, TRCAA accepts any OpenAI-API-compatible endpoint as a Custom provider — self-hosted gateways, internal LiteLLM proxies, vLLM/TGI servers, or anything else that speaks the/v1/chat/completions shape. Point it at a base URL and, if the endpoint requires one, an API key. This is the same mechanism the AWS Bedrock setup below uses.

AWS Bedrock via LiteLLM

A worked example of a Custom provider: for enterprise environments with an existing AWS contract, route Claude through Bedrock using a local LiteLLM proxy:

1. Install LiteLLM

pip install litellm[proxy]

2. Configure it

# ~/.litellm/config.yaml
model_list:
  - model_name: bedrock-claude
    litellm_params:
      model: bedrock/us.anthropic.claude-sonnet-4-6
      aws_region_name: us-east-1

general_settings:
  master_key: sk-your-secure-key

3. Run it

nohup litellm --config ~/.litellm/config.yaml --port 8000 > ~/.litellm/litellm.log 2>&1 &

4. Point TRCAA at it

Provider: OpenAI (OpenAI-compatible)
Base URL: http://localhost:8000/v1
API key: whatever you set as master_key
Model: bedrock-claude