OpenAI Provider¶

Integrates OpenAI's GPT models into Amplifier.

Module ID¶

provider-openai

Installation¶

providers:
  - module: provider-openai
    source: git+https://github.com/microsoft/amplifier-module-provider-openai@main

Configuration¶

Option	Type	Default	Description
`api_key`	string	`$OPENAI_API_KEY`	API key
`base_url`	string	null	Optional custom endpoint (`$OPENAI_BASE_URL`); null = OpenAI default
`default_model`	string	`gpt-5.4`	Default model
`max_tokens`	integer	`4096`	Maximum output tokens
`temperature`	float	null	Sampling temperature; null = not sent (some models don't support it)
`reasoning`	string	null	Reasoning effort: `minimal\\|low\\|medium\\|high`; null = not sent
`reasoning_summary`	string	`detailed`	Reasoning verbosity: `auto\\|concise\\|detailed`
`truncation`	string	`auto`	Automatic context management
`enable_state`	boolean	`false`	Enable stateful conversations
`debug`	boolean	`false`	Enable standard debug events
`raw_debug`	boolean	`false`	Enable ultra-verbose raw API I/O logging
`priority`	int	100	Provider priority for selection
`timeout`	float	600.0	API timeout in seconds
`max_retries`	int	5	Retry attempts before failing
`retry_jitter`	float	0.2	Randomness in retry delays (0.0-1.0)
`min_retry_delay`	float	1.0	Minimum delay between retries
`max_retry_delay`	float	60.0	Maximum delay between retries

Supported Models¶

Model	Description
`gpt-5.4`	GPT 5.4 (default)
`gpt-5.4-pro`	GPT 5.4 Pro
`gpt-5.1`	GPT 5.1
`gpt-5.1-codex`	GPT-5.1 codex
`gpt-5-mini`	Smaller, faster GPT-5
`gpt-5-nano`	Smallest GPT-5 variant
`o3-deep-research`	o3 Deep Research
`o4-mini-deep-research`	o4-mini Deep Research

Reasoning Configuration¶

Control the model's reasoning effort and output verbosity:

providers:
  - module: provider-openai
    config:
      reasoning: "medium"  # minimal|low|medium|high
      reasoning_summary: "detailed"  # auto|concise|detailed

Reasoning levels: - minimal - Quick responses, minimal reasoning - low - Light reasoning for simple tasks - medium - Balanced reasoning for most tasks - high - Deep reasoning for complex problems

Summary verbosity: - auto - Model decides appropriate detail level - concise - Brief reasoning summaries - detailed - Verbose reasoning output (similar to Anthropic's thinking blocks)

Stateful Conversations¶

Enable conversation persistence across requests:

providers:
  - module: provider-openai
    config:
      enable_state: true

With state enabled, the provider tracks conversation history server-side.

Truncation¶

Automatic context management to handle token limits:

providers:
  - module: provider-openai
    config:
      truncation: "auto"  # auto|none

When set to auto, OpenAI automatically truncates older messages to fit within context limits.

Tool Calling¶

The provider automatically detects and processes tool calls from the Responses API:

providers:
  - module: provider-openai
    config:
      default_model: gpt-5.4

Tools declared in your configuration are available to the model automatically.

Debug Events¶

Enable debug logging:

providers:
  - module: provider-openai
    config:
      debug: true  # Standard debug events
      raw_debug: true  # Ultra-verbose raw API I/O

Events: - llm:request:debug - Request summary - llm:response:debug - Response summary - llm:request:raw - Complete request params - llm:response:raw - Complete response object

Environment Variables¶

export OPENAI_API_KEY="your-api-key-here"
export OPENAI_BASE_URL="https://custom-endpoint.com"  # Optional

Example Configuration¶

providers:
  - module: provider-openai
    source: git+https://github.com/microsoft/amplifier-module-provider-openai@main
    config:
      api_key: ${OPENAI_API_KEY}
      default_model: gpt-5.4
      max_tokens: 4096
      temperature: 0.7
      reasoning: medium
      reasoning_summary: detailed
      truncation: auto