Sources

phoenix

Read transcripts from Arize Phoenix traces.

Each Phoenix trace (collection of spans with same trace_id) becomes one Scout transcript. Child spans (LLM calls, tools) become events within the transcript.

Source

async def phoenix(
    project: str | None = None,
    from_time: datetime | None = None,
    to_time: datetime | None = None,
    trace_id: str | list[str] | None = None,
    session_id: str | list[str] | None = None,
    tags: list[str] | None = None,
    metadata: dict[str, str] | None = None,
    limit: int | None = None,
    api_key: str | None = None,
    base_url: str | None = None,
) -> AsyncIterator[Transcript]

project str | None: Phoenix project name or ID. Required for fetching spans.
from_time datetime | None: Only fetch traces created on or after this time
to_time datetime | None: Only fetch traces created before this time
trace_id str | list[str] | None: Fetch specific trace(s) by ID (single string or list)
session_id str | list[str] | None: Filter to traces belonging to specific session(s) (matches session.id attribute on root span)
tags list[str] | None: Filter by tags on root span (all must match; stored in tag.tags attribute)
metadata dict[str, str] | None: Filter by metadata key-value pairs on root span (all must match; stored in metadata.* attributes)
limit int | None: Maximum number of transcripts to fetch
api_key str | None: Phoenix API key (or PHOENIX_API_KEY env var)
base_url str | None: Phoenix base URL (or PHOENIX_COLLECTOR_ENDPOINT env var)

langsmith

Read transcripts from LangSmith traces.

Each LangSmith root run (trace) becomes one Scout transcript. Child runs (LLM calls, tools) become events within the transcript.

Data sources:

project - Import traces from a project (default)
dataset - Import examples from a dataset

Source

async def langsmith(
    project: str | None = None,
    dataset: str | None = None,
    from_time: datetime | None = None,
    to_time: datetime | None = None,
    tags: list[str] | None = None,
    filter: str | None = None,
    trace_filter: str | None = None,
    run_type: str | None = None,
    limit: int | None = None,
    api_key: str | None = None,
    api_url: str | None = None,
) -> AsyncIterator[Transcript]

project str | None: LangSmith project name. Optional for dataset, used as default data source otherwise.
dataset str | None: LangSmith dataset name or ID. Fetches from curated evaluation datasets instead of project traces.
from_time datetime | None: Only fetch traces created on or after this time
to_time datetime | None: Only fetch traces created before this time
tags list[str] | None: Filter by tags (all must match)
filter str | None: LangSmith filter string for run filtering. Can use filters saved from the LangSmith UI.
trace_filter str | None: Filter on root run attributes
run_type str | None: Filter by run type (llm, chain, tool, etc.)
limit int | None: Maximum number of transcripts to fetch
api_key str | None: LangSmith API key (or LANGSMITH_API_KEY env var)
api_url str | None: LangSmith API URL (or LANGSMITH_ENDPOINT env var)

logfire

Read transcripts from Logfire traces.

Each Logfire trace (collection of spans with same trace_id) becomes one Scout transcript. Child spans (LLM calls, tools) become events within the transcript.

Source

async def logfire(
    project: str | None = None,
    from_time: datetime | None = None,
    to_time: datetime | None = None,
    filter: str | None = None,
    trace_id: str | None = None,
    limit: int | None = None,
    read_token: str | None = None,
) -> AsyncIterator[Transcript]

project str | None: Logfire project name in format “org/project”. If not provided, queries across all accessible projects.
from_time datetime | None: Only fetch traces created on or after this time
to_time datetime | None: Only fetch traces created before this time
filter str | None: SQL WHERE fragment for additional filtering (e.g., “attributes->>‘gen_ai.request.model’ = ‘gpt-4o’”)
trace_id str | None: Fetch a specific trace by ID instead of querying
limit int | None: Maximum number of transcripts to fetch
read_token str | None: Logfire read token (or LOGFIRE_READ_TOKEN env var). Generate from Logfire dashboard Settings > Read Tokens.

weave

Read transcripts from W&B Weave traces.

Each Weave trace (root call + children) becomes one Scout transcript. Child calls (LLM calls, tools) become events within the transcript.

Source

async def weave(
    project: str,
    from_time: datetime | None = None,
    to_time: datetime | None = None,
    filter: dict[str, Any] | None = None,
    limit: int | None = None,
    api_key: str | None = None,
) -> AsyncIterator[Transcript]

project str: W&B project name in format “entity/project”
from_time datetime | None: Only fetch traces created on or after this time
to_time datetime | None: Only fetch traces created before this time
filter dict[str, Any] | None: Dictionary of filters to apply to call query
limit int | None: Maximum number of transcripts to fetch
api_key str | None: W&B API key (or WANDB_API_KEY env var)

claude_code

Read transcripts from Claude Code sessions.

Each Claude Code session can contain multiple conversations separated by /clear commands. Each conversation becomes one Scout transcript.

When Claude Code enters plan mode and executes a plan, it creates separate session files that share the same slug. These related sessions are merged into a single transcript.

Source

async def claude_code(
    path: str | PathLike[str] | None = None,
    session_id: str | None = None,
    from_time: datetime | None = None,
    to_time: datetime | None = None,
    limit: int | None = None,
) -> AsyncIterator["Transcript"]

path str | PathLike[str] | None: Path to Claude Code project directory or specific session file. If None, scans all projects in ~/.claude/projects/
session_id str | None: Specific session UUID to import
from_time datetime | None: Only fetch sessions modified on or after this time
to_time datetime | None: Only fetch sessions modified before this time
limit int | None: Maximum number of transcripts to yield