Extended OpenAI Conversation (EOC)

A maintained fork of Extended OpenAI Conversation for Home Assistant that:

• Works with OpenAI Python SDK 1.x

• Supports reasoning‑class models via the Responses API (e.g., GPT‑5 family)

• Integrates cleanly with Assist pipelines

• Keeps memory scaffolding present but OFF by default

• Upstream: https://github.com/jekalmin/extended_openai_conversation

• This fork: https://github.com/Da3dalusCode/extended_openai_conversation

Features

• Assist‑compatible conversation entity (returns current ConversationResult)
• Responses API for reasoning models
  • Inputs as input_text items
  • System prompt in instructions
  • reasoning: { effort } when applicable
  • No temperature/top_p on reasoning endpoints
• Chat Completions for non‑reasoning models (sampling allowed)
• Options gear (model, strategy, effort, max tokens, temp/top‑p, prompt) via OptionsFlowWithReload
• Azure OpenAI support (Base URL + API version)
• Async‑safe client using HA’s shared HTTPX to avoid blocking SSL CA loads

Requirements

• Home Assistant 2025.4.0+ recommended
• Python OpenAI SDK >=1.0.0,<2.0.0 (installed by HA from manifest.json)
• HACS for installation

Install (HACS)

1. HACS → Custom repositories → add this repo (category Integration)
2. HACS → Download → Restart Home Assistant when prompted
3. Settings → Devices & Services → Add Integration → “Extended OpenAI Conversation”
4. Enter API key (set Base URL + API version for Azure)

Configure

Open the integration card → Configure (gear):

• Model (default gpt-5)
• Model strategy: auto | force_chat_completions | force_responses_api
• Use Responses API (for non‑reasoning models when strategy is auto)
• Reasoning effort: minimal | low | medium | high
• Max tokens, Temperature, Top‑p, System prompt

Recommended (long, smart chats)

• Model: gpt-5
• Strategy: auto
• Use Responses API: on
• Effort: medium (raise to high as needed)
• Max tokens: 800–1200

With Responses API, system text goes into instructions and inputs use input_text.

Assist usage

• Settings → Voice Assistants → set Conversation agent = Extended OpenAI Conversation
• Use Assist (text/voice) or ESPHome satellites normally.
• The agent returns continue_conversation when follow‑up is implied.

Troubleshooting

• 400: invalid input type (older builds): ensure you’re on v1.4.1+ (we use input_text + instructions). :contentReference[oaicite:8]{index=8}
• intent-failed with .as_dict: fixed in v1.4.1 (compat shim). :contentReference[oaicite:9]{index=9}
• Blocking SSL warning (load_verify_locations): fixed by using HA’s shared HTTPX client in v1.4.1. :contentReference[oaicite:10]{index=10}

Enable debug:

action: logger.set_level
data:
  custom_components.extended_openai_conversation: debug
  homeassistant.components.conversation: debug

Memory scaffolding

• Present but OFF by default (no network calls). When enabled in a future release, memory will write/query a provider and prepend retrieved context to instructions, following HA’s LLM guidance.