Configure Bring Your Own AI¶

QLens reads its provider registry from:

Linux/macOS: ~/.config/qlens/ai.yaml
Windows: %APPDATA%\qlens\ai.yaml
Override: QLENS_AI_CONFIG=/path/to/ai.yaml

Inspect built-in defaults¶

qlens ai list
qlens ai validate
qlens ai doctor mock

Native OpenAI¶

Create the provider without storing the key:

qlens ai init openai \
  --type openai \
  --model-env QLENS_AI_MODEL \
  --api-key-env OPENAI_API_KEY \
  --default

Then set credentials in the shell:

export OPENAI_API_KEY="your-secret-key"
export QLENS_AI_MODEL="your-approved-model"
qlens ai doctor openai

Generate a Bangla monologue:

qlens publication ai-draft PUB-000001 \
  --provider openai \
  --profile monologue_bn \
  --actor Mamun

OpenRouter or another OpenAI-compatible API¶

qlens ai init openrouter \
  --type openai_compatible \
  --base-url https://openrouter.ai/api/v1 \
  --api-key-env OPENROUTER_API_KEY \
  --model openai/gpt-5 \
  --default

export OPENROUTER_API_KEY="your-key"
qlens ai doctor openrouter

The same pattern works for a private compatible gateway by changing --base-url, key environment variable, and model.

Local Ollama¶

Install Ollama, pull a model, then register it:

ollama pull llama3.1
qlens ai init local-ollama \
  --type ollama \
  --base-url http://127.0.0.1:11434 \
  --model llama3.1 \
  --default
qlens ai doctor local-ollama --network

Generate locally:

qlens publication ai-draft PUB-000001 \
  --provider local-ollama \
  --profile monologue_en

Network diagnostics¶

Configuration-only diagnostics do not contact a provider:

qlens ai doctor openai

Explicit connection test:

qlens ai doctor openai --network

Security¶

Never write a key into ai.yaml, a publication object, .env committed to Git, NUR JavaScript, or command history. QLens rejects direct api_key, token, and secret fields.