Messages


POST https://api.fenno.ai/v1/messages
x-api-key: <FENNO_API_KEY>
anthropic-version: 2023-06-01
Content-Type: application/json

Request fields

Field	Type	Meaning
`model`	string, required	Exact Messages-capable model ID
`max_tokens`	integer, required	Output-token limit within the selected model’s bounds
`messages`	array, required	User/assistant history; content is a string or supported blocks
`system`	string or array, optional	System guidance outside the messages array
`stream`	boolean	Return message and content-block events
`tools`	array, optional	Definitions with `name`, `description` and `input_schema`
`tool_choice`, `thinking`	object, optional	Only supported values for the selected model

Python example

Install anthropic; set FENNO_API_KEY and a Messages-capable FENNO_MODEL:


import os
from anthropic import Anthropic
 
client = Anthropic(
    base_url="https://api.fenno.ai",
    api_key=os.environ["FENNO_API_KEY"],
    timeout=60.0,
    max_retries=0,
)
response = client.messages.create(
    model=os.environ["FENNO_MODEL"],
    max_tokens=512,
    system="Answer concisely.",
    messages=[{"role": "user", "content": "What is an API gateway?"}],
)
for block in response.content:
    if block.type == "text":
        print(block.text)
print(response.stop_reason)

Content and stop reasons

content[] is not always a single text block. Handle blocks by type: text, image/document input, tool_use, and supported thinking blocks are structurally different. Preserve model-required non-text blocks when carrying a conversation forward.

Common stop reasons include end_turn, max_tokens, stop_sequence and tool_use. A tool_use stop means the application should process allowed tool calls, not display them as a final answer. A max_tokens stop may leave output incomplete.

For a tool call, keep the assistant’s tool_use block and return a user message containing a tool_result with the matching tool_use_id. Do not place Chat Completions role: tool messages into this protocol.

Streaming lifecycle


message_start
content_block_start -> content_block_delta* -> content_block_stop
message_delta -> message_stop

There can be multiple content blocks, pings and error events. A delta can contain text, partial tool JSON or thinking data. Associate it with the block index and collect complete arguments before parsing.

The SDK’s client.messages.stream(...) context manager exposes text_stream for simple text use. Applications using tools must also process non-text content and the final stop reason.

See function calling, prompt caching, and error formats.