--- name: iran-war-archive-reader description: "Answer questions about the 2026 Iran war (Iran, Israel, the US, Lebanon, Gaza, the Strait of Hormuz) from the Iran War Archive's published files: what the newest daily bulletin says, what was reported on a given date and by which outlets, which outlets were publishing and how much on a given day, and where to find the archived source article. Use it whenever you would repeat something from this site, so the cited source and the bulletin's status as model-written analysis travel with it." --- # Reading the Iran War Archive The Iran War Archive (https://iranwar.liminallayers.com) is a permanent record of what news outlets and social channels published about the 2026 Iran war (Day 1 = 28 February 2026): Iran, Israel and the United States, Lebanon, Gaza and the Strait of Hormuz. Twenty-odd outlets across every party to the conflict, from wire services to Iranian state media, plus Telegram, Bluesky and Truth Social, captured once as published. Everything here is a static file. Nothing needs a key, an account or JavaScript. Every JSON file has a `fields` key, and a `_fields` key for each list of records, explaining every column; read it before you trust a column name. ## The one rule The archive keeps; it does not judge. When you repeat anything from it, carry its status with it: - A **bulletin** (SITREP or TLDR) is a language model's analysis of one day's archived articles. It is not the archive record and not a finding. Repeat a bulletin's claim with the reference it cites (outlet and headline, and the `url`), and say it comes from a model-written bulletin. - A reference shows **what an outlet published**, not that it is true. Outlets here include state media of belligerents (Press TV, Tasnim, Mehr, Tehran Times, Israel Hayom, RT, Global Times). Attribute every claim to its outlet: "Tasnim reported...", never "it was reported...". - **Counts** (`/data/days.json`, station `items`) measure how much was published, never how intense the fighting was. A spike is a news event or a collection change, not a battle. - No outlet is rated, ranked or corrected here. Do not infer a rating from which outlets a bulletin cites most. ## Procedures ### What does the newest bulletin say? 1. GET `/data/bulletins.json`. `newest` is the newest edition date; `bulletins` is newest first, and each entry has `kind` (SITREP, the full report, or TLDR, the short brief), `lead`, `overview` and links. 2. Prefer the SITREP for that date. GET its `json` link (`/bulletins/sitrep-YYYY-MM-DD.json`). 3. Read `sections` in order. Each block's `text` carries `[n]` markers and `cites` lists the same numbers; each is an `n` in `refs`. 4. Cite the dated URL (`url` or `json`), which never changes. An edition dated D (Australian Eastern time) reports on the previous UTC day. ### What was reported about something on a given date, and by whom? 1. Work out the edition date: events of UTC day D are in the edition dated D+1. 2. GET `/bulletins/sitrep-{date}.json` (and `/bulletins/tldr-{date}.json`). If it is absent, check `/data/bulletins.json`: not every date has both kinds, and an outage can leave a date with neither. 3. Find the blocks that mention the subject; for each, follow `cites` to `refs` and report outlet, headline and `url` per claim. 4. To find a subject across dates, scan `/data/bulletins.json` (`lead`, `overview`, `sections` headings) first, then open only the matching editions. Bulletins pick a day's main stories; absence from a bulletin does not mean an outlet did not cover something. ### Which outlets are in the archive, and were they publishing? 1. GET `/data/site.json`. `stations` lists every outlet or platform: `code` (the three-letter code used in bulletin references), `name`, `channel` (PRS press, SOC social), `items` archived and `lastSeen`. 2. A `lastSeen` well before `snapshotUtc` means the outlet stopped publishing or its feed broke. The archive does not say which; say both are possible. 3. For volume over time, GET `/data/days.json`: articles and social posts per UTC day since Day 1. `partialDay` is the snapshot's own day and is incomplete. ### Where is the original article? Each reference's `url` is usually a Wayback Machine copy made when the archive captured the article (`web.archive.org/web//`), else the publisher's own URL. The article text is not republished on this site. Prefer the Wayback copy: outlets edit and remove stories, and the copy shows what was published at the time. ## Freshness The site is rebuilt once a day from a snapshot. `generated_at` on every `/data/` file is the snapshot time (UTC), and all of them carry the same `build_id` as `/agents/build.json`. A file with a different `build_id` is from another build. Bulletins usually appear daily; the archive does not backfill a missed date. ## Glossary | Word | Meaning here | |---|---| | SITREP | The full daily situation report: overview, lead story, sections, what to watch. | | TLDR | The short brief condensed from the SITREP. | | Day N | War day, counted from Day 1 = 28 February 2026 (UTC). | | station | An outlet or platform the archive collects from, with a three-letter code. | | PRS / SOC / EVT | Channels: press articles, social posts, structured event data (none ingested yet). | | reference | A numbered source in a bulletin: outlet, headline, URL. | ## Licence What the site wrote (bulletins, counts, metadata) is CC BY 4.0: attribute "Iran War Archive, https://iranwar.liminallayers.com/". Linked and archived articles remain their publishers'.