OpenAI Integration

OpenAI API integration details

OpenAI Integration

Simili Bot v0.2.0 supports OpenAI as a full alternative to Google Gemini for both embeddings and LLM analysis.

Services

Simili Bot uses OpenAI for:

1. Text Embeddings - Convert text to vectors for semantic search
2. LLM Analysis - AI reasoning, duplicate detection, routing, and triage

Embeddings

Supported models

Model	Dimensions	Notes
text-embedding-3-small	1536	Default for OpenAI
text-embedding-3-large	3072	Highest quality
text-embedding-ada-002	1536	Legacy, not recommended

The dimensions value in your config must match the model’s output dimensions. Mismatches cause Qdrant collection errors.

LLM analysis

Default LLM model: gpt-5.2

All LLM tasks are supported with OpenAI:

• Duplicate detection
• Quality assessment
• Issue routing
• Auto triage

Provider precedence

If both GEMINI_API_KEY and OPENAI_API_KEY are set, Gemini takes priority.

To force OpenAI, either:

• Only set OPENAI_API_KEY (do not set GEMINI_API_KEY), or
• Explicitly set provider: "openai" in your config

Configuration

Embeddings only with OpenAI

embedding:
  provider: "openai"
  api_key: "${OPENAI_API_KEY}"
  model: "text-embedding-3-small"
  dimensions: 1536

Full OpenAI stack

embedding:
  provider: "openai"
  api_key: "${OPENAI_API_KEY}"
  model: "text-embedding-3-small"
  dimensions: 1536

llm:
  provider: "openai"
  api_key: "${OPENAI_API_KEY}"
  model: "gpt-5.2"

Mixed providers

Use Gemini for LLM and OpenAI for embeddings (or vice versa):

embedding:
  provider: "openai"
  api_key: "${OPENAI_API_KEY}"
  model: "text-embedding-3-large"
  dimensions: 3072

llm:
  provider: "gemini"
  api_key: "${GEMINI_API_KEY}"
  model: "gemini-2.5-flash"

Error handling

Simili Bot uses the same exponential backoff retry logic for OpenAI as for Gemini — typed error handling ensures transient failures are retried correctly.

Next steps

Gemini integration

Use Gemini instead

Environment variables

Configure API keys