Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/The-Context-Company/observatory/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

The run() function creates a new run to track the execution of your AI agent. A run represents a single interaction or task execution, containing the prompt, response, metadata, and child steps or tool calls.

​
Function Signature

from contextcompany import run

run(
    run_id: Optional[str] = None,
    session_id: Optional[str] = None,
    conversational: Optional[bool] = None,
    api_key: Optional[str] = None,
    tcc_url: Optional[str] = None,
) -> Run

​
Parameters

​
run_id
str
Unique identifier for this run. If not provided, a UUID will be automatically generated.
​
session_id
str
Session identifier to group multiple runs together (e.g., a conversation thread).
​
conversational
bool
Whether this run is part of a conversational flow. Set to True for chat-based interactions.
​
api_key
str
Observatory API key. If not provided, uses the TCC_API_KEY environment variable.
​
tcc_url
str
Custom Observatory endpoint URL. If not provided, uses the TCC_URL environment variable or defaults to production.

​
Returns

Returns a Run object with the following methods:

​
prompt()

Set the prompt for the run: 
r.prompt(
    user_prompt: str,
    system_prompt: Optional[str] = None
) -> Run
​
user_prompt
str
required
The user’s input or query
​
system_prompt
str
System instructions or context

​
response()

Set the final response: 
r.response(text: str) -> Run
​
text
str
required
The agent’s final response text

​
status()

Set the status code and optional message: 
r.status(
    code: int,
    message: Optional[str] = None
) -> Run
​
code
int
required
Status code: 0 = success, 1 = partial success, 2 = error
​
message
str
Human-readable status message

​
metadata()

Attach custom metadata: 
r.metadata(
    data: Optional[Dict[str, str]] = None,
    **kwargs: str
) -> Run
​
data
Dict[str, str]
Dictionary of metadata key-value pairs
​
**kwargs
str
Additional metadata as keyword arguments

​
step()

Create a child step: 
r.step(step_id: Optional[str] = None) -> Step
See step() for full documentation.

​
tool_call()

Create a child tool call: 
r.tool_call(
    tool_name: Optional[str] = None,
    tool_call_id: Optional[str] = None
) -> ToolCall
See tool_call() for full documentation.

​
feedback()

Submit user feedback for this run: 
r.feedback(
    score: Optional[Literal["thumbs_up", "thumbs_down"]] = None,
    text: Optional[str] = None
) -> bool
​
score
Literal['thumbs_up', 'thumbs_down']
Thumbs up or thumbs down rating
​
text
str
Textual feedback (max 2000 characters)

​
end()

Finalize and send the run data: 
r.end() -> None
You must call r.prompt() before calling r.end(), or a ValueError will be raised.

​
error()

Mark the run as failed and send immediately: 
r.error(status_message: str = "") -> None
​
status_message
str
Error message describing what went wrong

​
Properties

​
run_id

Get the run identifier: 
run_id: str = r.run_id

​
Usage Examples

​
Basic Usage

from contextcompany import run

r = run()
r.prompt(
    user_prompt="Summarize this article",
    system_prompt="You are a helpful assistant."
)
r.response("Here is a summary of the article...")
r.end()

​
Conversational Agent

from contextcompany import run
import uuid

session_id = str(uuid.uuid4())

# First message
r1 = run(session_id=session_id, conversational=True)
r1.prompt(user_prompt="Hello!")
r1.response("Hi! How can I help you?")
r1.end()

# Follow-up message
r2 = run(session_id=session_id, conversational=True)
r2.prompt(user_prompt="What's the weather?")
r2.response("Let me check the weather for you.")
r2.end()

​
With Metadata

from contextcompany import run

r = run()
r.prompt(user_prompt="Analyze sales data")
r.metadata(
    user_id="user_123",
    department="sales",
    region="us-west"
)
r.response("Sales analysis complete.")
r.end()

​
With Steps and Tool Calls

from contextcompany import run

r = run()

# Track an LLM call
s = r.step()
s.prompt("What's the weather in SF?")
s.response("I'll check the weather for you.")
s.model(requested="gpt-4", used="gpt-4")
s.tokens(prompt_uncached=20, completion=15)
s.end()

# Track a tool call
tc = r.tool_call(tool_name="get_weather")
tc.args({"location": "San Francisco"})
tc.result({"temperature": 65, "condition": "sunny"})
tc.end()

# Finalize the run
r.prompt(user_prompt="What's the weather in SF?")
r.response("It's 65°F and sunny in San Francisco.")
r.end()

​
Error Handling

from contextcompany import run

r = run()
r.prompt(user_prompt="Process this request")

try:
    # Your agent logic here
    result = process_request()
    r.response(result)
    r.end()
except Exception as e:
    r.error(f"Failed to process request: {str(e)}")

​
Custom Status Codes

from contextcompany import run

r = run()
r.prompt(user_prompt="Generate report")

try:
    report = generate_report()
    if report.is_complete:
        r.status(0, "Success")  # Complete success
    else:
        r.status(1, "Partial success")  # Partial success
    r.response(report.content)
    r.end()
except Exception as e:
    r.error(f"Report generation failed: {str(e)}")

​
Best Practices

  1. Always call end(): Ensure r.end() or r.error() is called to send the run data to Observatory.
  2. Set prompt before ending: The prompt is required before calling r.end().
  3. Use session_id for conversations: Group related runs together using the same session_id.
  4. Add meaningful metadata: Attach user IDs, request IDs, or other context to help with debugging and analysis.
  5. Handle errors gracefully: Use r.error() to capture and report failures.

​
See Also

Build docs developers (and LLMs) love

Get started for freeTalk to us