Daily opportunity-signal digest automation for Telegram

What it calls

These are the endpoints this build calls directly, with runtime values replaced by a placeholder.

• the Hacker News Algolia API, lobste.rs, Reddit, arXiv and the GitHub search API, all keyless
• the digest model call has no filesystem, shell or network access beyond localhost Ollama

Safety checks

An automation that writes into a system of record has to refuse some cases. These are the checks this build makes before it acts.

an optional local Ollama qwen3:14b

`sanitize.mjs` strips forged prompt-injection tags from every fetched item, keeping and labelling rather than dropping

nothing in the pipeline executes, clones or follows what it finds

Where it runs

It runs as a Windows scheduled task (daily 07:30).

It also touches Hacker News, Reddit, ArXiv, GitHub and Ollama.

What this record is

Data Jockey is a services firm, so this page is an implementation record. There is no product to buy here and no price on this page.

No customer is named on it either. Data Jockey publishes no customer list, and the calls and checks above are the part a reader can check.

Questions we hear

Is daily opportunity-signal digest automation for Telegram a product or a project?

It is a project. The queries on this page are specific to how one company had its Telegram set up: which fields carried the human identifier, which documents already existed, and where a human had to stay in the loop. A product cannot know any of that in advance, which is why the integration is the work.

Did this run in production?

Yes. It runs in production today as a Windows scheduled task (daily 07:30), and this page is read out of the code that runs.

What happens when the automation is not sure?

It stops. This build carries 3 distinct kinds of check, and every one of them exists to hand the case to a person. Automating the confident cases and routing the rest is what makes an integration safe to leave running.

Can Data Jockey build this for us?

Book a call and we will scope it against the systems you already run. Data Jockey publishes no prices, so there is no figure on this page, and what a build costs depends on what your setup turns out to need.

Where these facts come from

Every call, table, check and setting above comes from the delivery record of this build, and each identifier in it was checked against the file that implements it by tools/capture-proof-of-work.mjs. The capture is committed at data/raw/dj-proof-of-work-capture-2026-09-07.json and records a sha256 of each source file. Generated 2026-09-07. No figure here was rounded, averaged or estimated.