venice-responses
Use Venice's Alpha POST /responses endpoint - an OpenAI-compatible Responses API with typed output blocks (reasoning, message, function_call, web_search_call). Covers request shape, streaming, differences from /chat/completions, supported venice_parameters subset, and E2EE behavior.
veniceai · 已有 AI 审核记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add veniceai/skills --skill venice-responses技能说明
name: venice-responses description: Use Venice's Alpha POST /responses endpoint - an OpenAI-compatible Responses API with typed output blocks (reasoning, message, function_call, web_search_call). Covers request shape, streaming, differences from /chat/completions, supported venice_parameters subset, and E2EE behavior.
Venice Responses API (Alpha)
POST /api/v1/responses is Venice's OpenAI-compatible Responses endpoint. It returns a structured, typed output array instead of a single message.content string — ideal for agents that need to separate reasoning, messages, tool calls, and built-in tool events.
Alpha. Access is gated behind the
responsesApiEnabledflag on Bearer API keys (staff-only during beta). x402 wallet auth bypasses this flag — you can pay per request without the flag. Schemas may change.
Use when
- You need the OpenAI Responses-style response shape (
output[]with typedtype: "reasoning" | "message" | "function_call" | "web_search_call"blocks) for a client library that expects it. - You want clean separation of reasoning vs message vs tool-call output.
- You want streaming via SSE with typed events.
Otherwise use venice-chat — it has more features, more models, and full Venice parameters.
Limitations vs /chat/completions
| Limitation | Detail |
|---|---|
| Stateless | No conversation persistence across requests. Send the full history each call. |
| E2EE models default to rejection | E2EE-capable models return 400 unless you pass venice_parameters.enable_e2ee: false (TEE-only mode). For end-to-end encrypted inference with E2EE headers, use /chat/completions. |
Subset of venice_parameters | character_slug, enable_e2ee, enable_web_search, enable_web_scraping, enable_web_citations, include_venice_system_prompt, include_search_results_in_stream are supported. strip_thinking_response, disable_thinking, enable_x_search are not wired through in Alpha. |
| Access gated by feature flag | Bearer keys without responsesApiEnabled get 401. x402 requests are allowed (pay-per-call). |
Authentication
Same as the rest of the API — either Authorization: Bearer <key> or SIGN-IN-WITH-X: <SIWX>. See venice-auth.
Minimal request
curl https://api.venice.ai/api/v1/responses \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "zai-org-glm-5-1",
"input": "Explain why the sky is blue in one paragraph."
}'
input accepts:
- a plain string, or
- an array of typed input items (similar to
chat/completionsmessage parts) for multi-turn or multimodal history.
Response shape
{
"id": "resp_abc123",
"object": "response",
"created_at": 1735689600,
"model": "zai-org-glm-5-1",
"status": "completed",
"output": [
{
"type": "reasoning",
"id": "rs_1",
"summary": ["I considered Rayleigh scattering..."],
"encrypted_content": "..."
},
{
"type": "message",
"id": "msg_1",
"status": "completed",
"role": "assistant",
"content": [{
"type": "output_text",
"text": "The sky is blue because...",
"annotations": [{
"type": "url_citation",
"url": "https://example.com/rayleigh",
"title": "Rayleigh scattering",
"start_index": 42,
"end_index": 99
}]
}]
},
{
"type": "function_call",
"id": "fc_1",
"call_id": "call_abc",
"name": "get_weather",
"arguments": "{\"city\":\"Paris\"}",
"status": "completed"
},
{
"type": "web_search_call",
"id": "ws_1",
"status": "completed"
}
],
"usage": {
"input_tokens": 20,
"input_tokens_details": {"cached_tokens": 0},
"output_tokens": 80,
"output_tokens_details": {"reasoning_tokens": 40},
"total_tokens": 100
}
}
Top-level status ∈ completed | failed | in_progress | cancelled. On failed, error.code and error.message are populated.
Output block types
type | Purpose |
|---|---|
reasoning | Thought process from reasoning models. summary[] holds human-readable text; encrypted_content holds opaque signatures — round-trip verbatim for multi-turn tool calls. |
message | Main text output. content[].type === "output_text", plus annotations[] for url_citation entries from web search. |
function_call | Tool call: name, stringified-JSON arguments, call_id. |
web_search_call | Sentinel showing the built-in web_search tool fired; use alongside url_citation annotations on messages. |
Match tool outputs back by call_id when continuing the turn.
Common request fields
| Field | Notes |
|---|---|
model | Required. Model ID, trait, or compatibility mapping. Feature suffixes allowed (see venice-chat). |
input | Required. String or input-items array. To set system/developer context, include a leading message with role: "system"/"developer" in the input array. |
tools | Array of {type:"function",function:{...}} or built-in {type:"web_search"} — availability depends on the model. |
tool_choice | "auto" / "required" / "none" / {type:"function",function:{"name":"..."}}. |
reasoning.effort | Reasoning effort hint for thinking models ("low" | "medium" | "high"). Mapped to reasoning_effort. |
temperature, top_p, max_output_tokens | Standard generation controls. max_output_tokens maps to max_tokens. |
web_search | Boolean shortcut for enabling web search, equivalent to adding {"type":"web_search"} to tools or setting venice_parameters.enable_web_search. |
include | Array of additional response fields to include (OpenAI compat). |
fallbacks | Up to 10 entries. Anthropic beta parameter for Claude Fable 5 server-side refusa |