Skip to main content
POST
/
v1
/
chat
/
completions
Chat Completions
curl --request POST \
  --url https://api.example.com/v1/chat/completions \
  --header 'Authorization: <authorization>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "model": "<string>",
  "messages": [
    {}
  ],
  "stream": true,
  "temperature": 123,
  "max_tokens": 123,
  "nightly": [
    {}
  ],
  "guardrail_policy": "<string>",
  "chunk_size": 123,
  "sliding_window": 123,
  "finetune_thresh": 123,
  "min_finetune_group": 123,
  "lora_name": "<string>",
  "adaption_inference": true
}
'
{
  "id": "<string>",
  "object": "<string>",
  "choices": [
    {}
  ],
  "usage": {}
}

​
POST /v1/chat/completions

Route chat completion requests to OpenRouter with optional Nightly memory, Policies guardrails, and Adaption fine-tuning support. Compatible with the OpenAI SDK.

​
Authentication

​
Authorization
string
required
Bearer token with your API key. Format: Bearer YOUR_ASYMMETRIC_API_KEY

​
Request Body

​
model
string
required
The model to use (e.g., openai/gpt-4o-mini, anthropic/claude-3-sonnet, etc.)
​
messages
array
required
Array of message objects with role and content fields
​
stream
boolean
default:"false"
Whether to stream the response
​
temperature
number
Sampling temperature (0-2)
​
max_tokens
integer
Maximum tokens to generate

​
Nightly (Memory) Parameters

​
nightly
array
Enable agent memory. Format: ["memory_group", "user_goal"]
  • memory_group: Unique identifier for the memory pool
  • user_goal: Description to help curate relevant memories

​
Policies (Guardrails) Parameters

​
guardrail_policy
string
Natural language policy to check outputs against. If provided, outputs are filtered in real-time.
​
chunk_size
integer
default:"10"
For streaming: how often to check guardrails (every N chunks)
​
sliding_window
integer
default:"5"
For streaming: number of recent chunks to check together

​
Adaption (Fine-tuning) Parameters

​
finetune_thresh
integer
Memory call count threshold to trigger training queue
​
min_finetune_group
integer
default:"5"
Minimum memories needed before triggering training
​
lora_name
string
Name of the LoRA adapter to use or create
​
adaption_inference
boolean
default:"false"
Set to true to use your finetuned LoRA adapter for inference

​
Response

Returns an OpenAI-compatible chat completion response.
​
id
string
Unique identifier for the completion
​
object
string
Always chat.completion
​
choices
array
Array of completion choices
​
usage
object
Token usage statistics

​
Example

from openai import OpenAI

client = OpenAI(
    base_url="https://rkdune--symmetry.modal.run/v1/",
    api_key="YOUR_ASYMMETRIC_API_KEY",
)

# Basic completion with memory and guardrails
completion = client.chat.completions.create(
    model="openai/gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello!"}],
    extra_body={
        "nightly": ["my_agent", "A helpful assistant"],
        "guardrail_policy": "Flag any inappropriate content"
    }
)

print(completion.choices[0].message.content)

​
Errors

StatusDescription
400Guardrail violation detected or invalid parameters
401Invalid API key
402Insufficient credits