Responses API

Use Responses only with a model that declares openai.responses. Supporting Chat Completions does not imply Responses support. Query GET /openai/v1/models?api=responses; on October 13, 2026 it returned deepseek-flash and glm-5.3-flash.

1. Prepare your terminal

Use Python 3.10+ and uv. Put your Praxis cloud API Key in PRAXIS_API_KEY; do not put it in source code or logs.

test -n "$PRAXIS_API_KEY"

Expected: exit code 0. Exit code 1 means the variable is missing. Examples default to https://api.deepraxis.com. To switch environments, set PRAXIS_API_BASE to the API origin without a trailing slash or API path.

2. Continue from the previous response

The platform expands the conversation history referenced by previous_response_id. The second call sends only the new question and the first response ID. Both calls use the same Key and model; the script checks that the second answer recalls a code absent from its new input.

Save as responses.py and run uv run -q --no-project --with openai --with anthropic --with google-genai python responses.py.

import os
from openai import OpenAI

client = OpenAI(api_key=os.environ['PRAXIS_API_KEY'],
    base_url=os.environ.get('PRAXIS_API_BASE', 'https://api.deepraxis.com') + '/openai/v1')
models = client.models.list(extra_query={'api': 'responses'})
print('X-Request-Id:', models._request_id)
assert 'deepseek-flash' in [model.id for model in models]
first = client.responses.create(model='deepseek-flash',
    input='Remember this code: LIME-742. Reply OK.',
    reasoning={'effort': 'none'}, max_output_tokens=512)
print('X-Request-Id:', first._request_id)
assert first.status == 'completed' and first.id
second = client.responses.create(model='deepseek-flash',
    previous_response_id=first.id, input='What was the code? Reply with only the code.',
    reasoning={'effort': 'none'}, max_output_tokens=512)
print('X-Request-Id:', second._request_id)
assert second.status == 'completed' and 'LIME-742' in second.output_text
assert second.usage is not None
print('answer:', second.output_text)
print('usage:', second.usage.model_dump())

Expected: both responses have status completed; the second answer contains LIME-742 and prints usage. Exit code 0. Reusing history is not free input: inspect usage for each turn.

Limits and observed failures

POST /openai/v1/responses/compact and POST /openai/v1/responses/input_tokens currently return HTTP 501 with not_implemented. Do not depend on them for conversation compaction or token preflight. Manage conversation length in your application and read actual usage after completion.

A request using gpt-5.6-sol returned HTTP 404 with model_protocol_not_offered. Choose a model from the Responses-filtered catalog. If the catalog assertion fails, recheck that catalog before retrying. Response IDs from another gateway do not move with your base URL; rebuild the necessary history.

See create a response for parameters and the OpenAI API reference for each operation's support status.

API reference → · Cookbook →