Sign in to Traq
Sign in with your Cross Call account. This browser remembers your access token until it expires or you sign out.
Console
Use the menu to open one action per page. Responses include method, URL, HTTP status, duration, request body, and response body.
Health Check
Checks whether the API server is alive.
System Stats
Shows MongoDB counts and Qdrant collection information.
External API Calls
Calls are grouped by the caller’s group ID. V3 searches assign one group to all provider requests they make; group totals add their token usage, estimated costs, and provider durations. The group time is the first API call in the group. Full payloads may include transcript content.
| First API call | Call group | Requests | Providers / models | Operations | Status | Tokens in / out | Total cost | Provider duration | Calls and details |
|---|---|---|---|---|---|---|---|---|---|
| Load calls to view the call log. | |||||||||
Database Records and Vector Bytes
Shows MongoDB record counts, Mongo storage bytes, Qdrant vector count, and actual vector bytes. Vector bytes are calculated as points_count * 3072 * 4.
Vectorize Call
Vectorize Raw Call
Paste a call with header lines like Workspace_ID, Call_ID, User_ID, Date, Time, Title, followed by timestamped transcript lines. The API converts it to the vectorization JSON and queues semantic vectorization.
Vectorize Records
Paste JSON shaped as {"output":[...]} or {"records":[...]}. Only text is embedded; every other field is stored as metadata.
Import Call
Paste a call with header lines like Workspace_ID, Call_ID, User_ID, Date, Time, Title, followed by a blank line and timestamped transcript lines.
Search transcript evidence
Choose a workspace and search method. Results are limited to calls you are authorized to read.
Answer
Multi Call Analyzer
Ask across selected calls using L1/L2/L3 context. History is sent as memory and updated after every answer.
V3 test suite
Import a Markdown query key, choose a workspace and one model, then run its questions across your authorized calls. Compare each answer with its grading reference and inspect the query details.
Run configuration
The suite runs one query at a time for the selected model.
No file imported. Built-in tests.txt will run.
Ready to run.
Results and grading
Use the filters to focus on completed answers, errors, or cases that still need a score.
Explore calls, chunks, and vectors
Choose a workspace, search its calls, and inspect every chunk with its metadata and saved vector values.
Workspaces
Select a workspace to browse its calls and search its chunks.
Edit Call
Delete Call
Delete Workspace
Deletes workspace calls, chunks, and related Qdrant vectors.