specstream-client

pip install specstream-client (the specstream package on PyPI is a different project)

Python client and command line for SpecStream, the live web data engine for teams and agents. Describe the list you need in plain words; SpecStream reads the web where those things live, hands back only the matches, and you pay for each result delivered.

Source URL and read time on every value.

The client has no dependencies and needs Python 3.12 or newer. It uses the same account, permissions, projects and balance as the workspace.

pip install specstream-client

The package installs the import name specstream and the command specstream.

Create an API key under Agents & API in your SpecStream workspace and keep it in the environment:

export SPECSTREAM_API_KEY="ss_live_..."

Read a request, get the price

import os
from specstream_client import Client

client = Client(os.environ["SPECSTREAM_API_KEY"])

read = client.compile("Open-source document-signing tools I can self-host. Include the license and latest release.")
offer = client.quote_capacity_work({
    "ast_hash": read["ast_hash"],
    "operation": "find",
    "requested_units": read["ast"]["limit"]["max"],
})
print(offer["maximum_cents"], offer["expires_at"])

Nothing has run yet. Review the quote: what will be found, the facts that come back, the most the work can cost, and when the price expires. A price lasts 30 minutes.

Approve it, then read the results

After a person approves, send the approved maximum in cents and a request key you saved first:

run = client.authorize_capacity_work({
    "quote_id": offer["quote_id"],
    "quote_hash": offer["quote_hash"],
    "confirm_max_cents": approved_maximum_cents,
    "idempotency_key": saved_request_key,
})
find_id = run["find_id"]

page = client.rows(find_id)
for row in page["rows"]:
    print(row)

If the response is lost, get_capacity_request(saved_request_key) returns the same work. Keep the original key when you retry, and you will not pay twice. get_find reports progress, and rows pages through results with next_cursor. Reading, exporting and sharing results you own cost nothing.

authorize_capacity_work refuses to send anything unless the quote, its signature, the maximum and the key are all present.

More than a first request

• quote_capacity_work also prices extend (more results), add_facts and refresh (a recheck, at half price) on a finished run. Pass the original parent_find_id.
• spec_options(kind) lists the facts you can ask for, per kind of thing. edit_spec changes a request without rereading it.
• create_capacity_schedule keeps a request current, hourly to yearly, inside per-run and per-period limits.
• export and download_export give CSV, XLSX, JSON or HTML. create_customer_webhook sends signed events.
• Projects, members, shares and starred items have their own methods.

Kinds of thing include local businesses, companies, people, creators, products, apps, job postings, papers, clinical trials, grants, tenders, regulations, vulnerabilities, software packages, repositories, datasets, books and podcasts.

Command line

specstream compile "Food banks in Texas, with their annual revenue."
specstream get-capacity-balance
specstream quote-capacity-work --body quote-request.json
specstream authorize-capacity-work --body approved-execution.json
specstream rows --find-id FIND_ID

--body takes a JSON file or - for stdin. Every command's --help describes its parameters. The command line exposes the same operations as the client.

Pricing

New accounts start with $10 of work for 30 days, no card. A delivered result costs 10 cents, 30 cents or $1 depending on how many facts it carries. No result, no charge. Rechecks cost half. People and agents share one balance. See pricing.

Links

• SDK docs and API docs
• Source and issues
• TypeScript client
• Support: support@specstream.ai

MIT license.