Upload AI Conversation Transcripts to a Project

Uploading transcript files is the fastest way to get conversation data into a Bigspin project. You can upload files when you first create a project, or add more transcripts to an existing project at any time. Bigspin supports several common file formats and parses them automatically so you don’t need to do any pre-processing before uploading.

Supported file formats

Bigspin accepts the following formats:

Format	Extension
Comma-separated values	`.csv`
Tab-separated values	`.tsv`
JSON array or object	`.json`
JSON Lines (one object per line)	`.jsonl`
Apache Parquet	`.parquet`

The maximum file size per upload is 500 MB. If your dataset exceeds this limit, split it into multiple files and upload them together — Bigspin accepts multiple files in a single upload.

Expected schemas

Bigspin is flexible about column and field naming, but each transcript must contain an identifier and the conversation content.

CSV / TSV
JSON / JSONL
Parquet

Each row in the file represents one transcript. The following columns are recognized:Required:

id — a unique identifier for the transcript
conversation (or messages, turns, or transcript) — the conversation content

Optional:

name — a human-readable label for the transcript
timestamp (or started_at or created_at) — when the conversation took place (ISO 8601 format recommended)

id,name,conversation,started_at
t1,Support chat,"[{""role"":""user"",""content"":""Hi""}]",2026-03-23T04:59:52Z

Each object represents one transcript. The following fields are recognized:Required:

id — a unique identifier for the transcript
turns — an array of turn objects, each with role and content

Optional:

name — a human-readable label for the transcript
started_at (or created_at) — when the conversation took place

Valid roles for turns: user, assistant, human_agent, system, toolExample JSONL (one object per line):

{"id": "t1", "turns": [{"role": "user", "content": "Hi"}, {"role": "assistant", "content": "Hello"}], "started_at": "2026-03-23T04:59:52Z"}
{"id": "t2", "turns": [{"role": "user", "content": "What can you do?"}, {"role": "assistant", "content": "I can help with many things."}], "started_at": "2026-03-23T05:12:00Z"}

Example JSON array:

[
  {
    "id": "t1",
    "turns": [
      {"role": "user", "content": "Hi"},
      {"role": "assistant", "content": "Hello"}
    ],
    "started_at": "2026-03-23T04:59:52Z"
  }
]

Parquet files follow the same column conventions as CSV/TSV. Bigspin reads the schema from the file and maps recognized column names automatically.Required columns: id and one of conversation, messages, turns, or transcriptOptional columns: name, timestamp, started_at, or created_at

How to upload transcripts

You can upload transcripts when creating a new project or from inside an existing project.

During project creation

Open the Create Project dialog

From the Projects dashboard, click Create Project.

Drag and drop or browse for files

In the Upload transcripts section of the dialog, drag your files onto the drop zone, or click the drop zone to open a file picker.You can select multiple files at once. Files that fail validation (wrong format or over 500 MB) are rejected individually — valid files in the same selection are still added.

Review your file list

Each added file appears below the drop zone with its name and size. Click the × next to a file to remove it before submitting.

Create the project

Click Create Project. Bigspin uploads and processes all selected files and takes you to the import progress screen.

In an existing project

Open the project

Click the project card on the Projects dashboard to open it.

Click the import button

Inside the project, click the Upload or import button to open the file upload interface.

Select your files

Drag files onto the drop zone or click to browse. Bigspin begins processing as soon as the upload completes.

What happens after upload

Once you upload files, Bigspin:

Validates each file’s format and size
Parses the transcripts according to the detected schema
Ingests the conversations into your project
Makes them available for analysis and reporting

If a file contains columns or fields that Bigspin doesn’t recognize, it ignores the unknown fields and processes everything else. Check the import progress screen for any parsing warnings.

Transcript Filters (with a connected data source)

If your workspace has a data source integration (such as LangSmith) connected, a Transcript Filters section appears in the Create Project dialog. Filters let you scope your project to only include transcripts whose metadata matches specific key/value criteria — for example, only traces tagged with a particular environment or user segment. As you add filter conditions, Bigspin shows a live preview count of how many transcripts currently match. Filters are applied at the subscription level and continue to be enforced on new traces as they sync in.

Transcript Filters are only available when a data source integration is connected to your workspace. See Connect Data Sources for setup instructions.

Documentation Index

​Supported file formats

​Expected schemas

​How to upload transcripts

​During project creation

​In an existing project

​What happens after upload

​Transcript Filters (with a connected data source)

Supported file formats

Expected schemas

How to upload transcripts

During project creation

In an existing project

What happens after upload

Transcript Filters (with a connected data source)