Senate on GitHub, Chamber SPARQL-only: two opposite ways to open data
Direct answer: the Italian Chamber and Senate both publish their data as open data, but with opposite strategies. The Chamber exposes everything only through a SPARQL endpoint (dati.camera.it/sparql, on Virtuoso). The Senate complements its SPARQL with a GitHub repository — SenatoDellaRepubblica/AkomaNtosoBulkData — with bills and amendments in Akoma Ntoso, CC BY 4.0, CI-updated. Integrating both for Open·Parlamento, we felt firsthand why this difference matters.
TL;DR
- Chamber: data via SPARQL only. No public versioned dumps. If the endpoint is slow or down, you have no plan B.
- Senate: SPARQL + bulk data on GitHub (AKN, CC BY 4.0, CI). You can clone, version and process the texts offline.
- Fragile «official» SPARQL: more than one Italian public endpoint answers
200but empty from a script — you must fall back to raw dumps. - Lesson: a versioned bulk dump is more robust than a live endpoint for anyone building data pipelines.
Why two different approaches?
The Chamber and Senate are autonomous institutions that opened their data at different times and with different technical choices. The Chamber invested in a SPARQL triple store (Virtuoso 07.10.3207): elegant for ad-hoc queries, but a single point of failure — if the endpoint slows down or returns empty, there’s no official alternative.
The Senate made a more «developer-minded» choice: besides SPARQL, it puts legislative texts in a GitHub repository updated via continuous integration. It’s a pattern data people know well: a versioned bulk dump is easy to clone, batch-process and reproduce, and doesn’t depend on a query server’s uptime.
| Aspect | Chamber | Senate |
|---|---|---|
| SPARQL endpoint | ✅ dati.camera.it/sparql |
✅ dati.senato.it/sparql |
| Public bulk dump | ❌ | ✅ GitHub (AKN, CC BY 4.0) |
| Versioned / CI | ❌ | ✅ CI-updated |
| Offline / batch work | hard | easy |
| Plan B if the endpoint is down | none | the dump |
The problem of «official» SPARQL endpoints that answer empty
Building the connectors, we found a recurring pattern in Italian public data: official SPARQL endpoints that, from a script, answer HTTP 200 with zero results, even though they work from a browser. It happens, for example, with the Constitutional Court’s open-data portal: «the official SPARQL answers 200 empty from a script → unusable», and the fix was to use the JSON datasets instead of the live endpoint.
This is exactly why the Senate’s choice — also publishing a versioned dump — is valuable: when the live endpoint betrays you, you still have a copy of the data you can process. Whoever publishes only SPARQL leaves you without a safety net.
What it means for anyone building on this data
- A SPARQL endpoint isn’t enough. Great for exploratory queries, fragile as the only access path. Serious pipelines need a dump.
- Versioning is a feature, not a detail. Knowing which version of the data you processed (a GitHub commit) makes analyses reproducible. A live endpoint doesn’t guarantee that.
- Normalize the diversity. Chamber via SPARQL, Senate via SPARQL + dump: our connector layer reads both and maps them to a single model before the knowledge graph. The graph is the door, but the engine is the normalized data asset behind it: we also covered the two ontologies OCD and OSR and why the real asset isn’t the graph visualization.
FAQ
Does the Chamber have a downloadable data dump?
As far as we found, no: the Chamber exposes its data publicly only through the SPARQL endpoint dati.camera.it/sparql. There’s no public versioned-dump repository equivalent to the Senate’s.
What is the Senate’s AkomaNtosoBulkData?
It’s an official GitHub repository of the Senate of the Republic with legislative texts (bills, amendments) in Akoma Ntoso format, under a CC BY 4.0 license and updated via continuous integration. It’s especially useful for amendment texts.
Why is a dump more reliable than a SPARQL endpoint?
A versioned dump is cloned once and processed offline: it doesn’t depend on server uptime, it’s reproducible (you know which version you used) and it doesn’t suffer timeouts or rate limits. A live endpoint is handy for ad-hoc queries but becomes a bottleneck for pipelines.
We build AI agents and knowledge graphs on real data — handling source diversity instead of hiding it. See how Open·Parlamento works, discover how agents and open data help you understand a country, or let’s talk.
Related articles
- Chatbots and the AI Act: What Really Lands on 2 August 2026From 2 August 2026 Article 50 of the EU AI Act applies: your chatbot must disclose it is an AI and generated content must be marked. Fines up to €15M or 3% of global turnover.
- Agentic Commerce: When an AI Agent Is the One Buying From YouChatGPT and Google opened checkout inside the AI. Your next customer may never see your site — only whether your catalog is machine-readable. Here is what to fix today.
- llms.txt Does (Almost) Nothing: The Data, and Why I Recommended It97% of sites with a valid llms.txt got zero requests for the file. Google says no Search system reads it. Here is what to do instead — and who actually does use it.