Delivery

We provision read access to Google Cloud Storage (GCS) buckets. You authenticate with a service account we provide and pull data on your own schedule.

Delivery Model

Each delivery is a delta. A timestamped folder contains only the signals detected since the previous delivery — not a cumulative snapshot. To build historical coverage, union folders backwards.

Past folders can receive corrections. If we fix data for a past period, we may update that folder in place. To pick up corrections, re-read recent folders.

Bucket versions (v1, v2, v3) are schema evolutions. The underlying signal data is the same — field names, types, or structure changed between versions. Looking far enough back may require reading from an older bucket version with a different schema. We're standardizing schemas across all signal types in 2026.

Backfill folders are historical deltas. Delivered at onboarding to provide coverage for time periods before your contract started.

Available Buckets

Each signal type has its own dedicated bucket. Access each bucket directly by its URI.

Financial

Social

Workforce

Market

Reference Data

Manifests cover most signal types today. We are adding manifests for HackerNews, ProductHunt, Federal Contract Awards, New Business Formation, Podcast, Patents and Reddit V3.

Bucket Structure

Each bucket contains timestamped folders. Each folder has two files: output.jsonl and output.parquet.

gs://autobound-news-v4/
├── 2026-08-27-09-04-12/
│   ├── output.jsonl
│   └── output.parquet
├── 2026-08-26-09-04-50/
│   ├── output.jsonl
│   └── output.parquet
└── ...

Pull from the most recent folder to get the latest data.

Authentication

1. We provide you with a GCP service account JSON key file
2. Authenticate: gcloud auth login --cred-file=/path/to/key.json
3. Use gcloud storage to access the buckets

# List folders in a bucket (access each bucket directly by URI)
gcloud storage ls gs://autobound-news-v4/

# Download JSONL
gcloud storage cp gs://autobound-news-v4/2026-08-27-09-04-12/output.jsonl ./

# Download Parquet
gcloud storage cp gs://autobound-news-v4/2026-08-27-09-04-12/output.parquet ./

📘

Tip: You can discover all available signal buckets by running gcloud storage ls --project=autobound-signal-delivery. This returns the names of all signal buckets in the project — including buckets you are not licensed for. You can browse bucket names and folder timestamps freely, but reading data inside a bucket requires objectViewer on that specific bucket, which is granted during onboarding. Unlicensed buckets return AccessDeniedException: 403 when you attempt to read their contents.

File Formats

Both formats contain the same data. Choose based on your pipeline.

{"signal_id":"7dfdb4b4-c0b4-4620-aca6-e7263123028e","signal_type":"news","detected_at":"2026-01-15T10:30:00Z","association":"company","company":{"name":"Acme Corp","domain":"acme.com"},"data":{"summary":"Acme Corp announces expansion into European markets...","source_url":"https://example.com/news/acme-expansion"}}

signal_id: STRING
signal_type: STRING
signal_subtype: STRING
detected_at: TIMESTAMP
association: STRING
company: STRUCT<name, domain, linkedin_url, industries, employee_count_low, employee_count_high, description>
contact: STRUCT<first_name, last_name, full_name, email, job_title, linkedin_url, city, state, country>
data: STRING (JSON-encoded)

Refresh Cadence

Getting Started

1. Contact [email protected] to get your service account credentials
2. Choose which signal types you need
3. Set up authentication and start pulling data