Bulk Parquet export
Download tweets and profiles for bulk analysis. Each daily export checks current membership and opt-outs before publication.
Download the daily data export, query public records through the API, or give an agent one canonical starting point.
The Supabase REST API is retired
The public Supabase REST API was retired on 19 November 2026: the tweet corpus no longer answers the anon key. Use the keyed Community Archive API at https://analytics.community-archive.org/v1 for filtered queries and the daily export for bulk data.
Point an agent here
https://www.community-archive.org/llms.txtThis plain-text index includes the data dump, API v1 routes and keys, query examples, schema notes, and deeper documentation links.
If you use the archive, please cite it
This applies to the export, the API and anything an agent writes from them. Use this citation, with the date you accessed the data:
Carvalho, F. (2024). Community Archive [Data set]. https://www.community-archive.org (accessed [date you accessed it])
Choose the policy-safe surface that fits your query.
Download tweets and profiles for bulk analysis. Each daily export checks current membership and opt-outs before publication.
Read tweets, timelines, search, accounts, members, likes and follows with a personal key. Cursor-paged, rate-limited, and served from the analytical store.
Signed-in owners can download their own processed archive JSON while current PostgreSQL policy permits it.
https://www.community-archive.org/api/archive/<username>
The daily package contains enriched tweets.parquet, separate profiles.parquet, and a manifest.json with row counts, schemas, and checksums. It includes eligible members and applies current consent to referenced authors too.
Use the latest release link below to find the most recent successful export. GitHub lists the download links; the files stay in consent-managed storage. Superseded packages are removed, so bookmark the latest release rather than an individual Parquet URL. Downloads may be temporarily withdrawn when consent changes.
Sign in, create a key on the API keys page, and send it as Authorization: Bearer cak_... on every request to https://analytics.community-archive.org/v1. Keep the key on your server or in your script, never in browser code. The full contract is the OpenAPI document.
export CA_API_KEY='cak_...' # create one at https://www.community-archive.org/settings/api-keys
# One account's tweets, newest first (page with next_cursor until it is null)
curl --get "https://analytics.community-archive.org/v1/accounts/exgenesis/tweets" \
-H "Authorization: Bearer $CA_API_KEY" \
--data-urlencode "order=desc" \
--data-urlencode "limit=5"
# Full-text search
curl --get "https://analytics.community-archive.org/v1/tweets/search" \
-H "Authorization: Bearer $CA_API_KEY" \
--data-urlencode "q=community archive" \
--data-urlencode "limit=5"const API = 'https://analytics.community-archive.org/v1'
const headers = { Authorization: `Bearer ${process.env.CA_API_KEY}` }
// Every tweet from one account, following the cursor
let cursor = null
do {
const url = new URL(`${API}/accounts/exgenesis/tweets`)
url.searchParams.set('order', 'asc')
url.searchParams.set('limit', '200')
if (cursor) url.searchParams.set('cursor', cursor)
const response = await fetch(url, { headers })
if (response.status === 429 || response.status === 503) {
const wait = Number(response.headers.get('Retry-After') ?? '5')
await new Promise((resolve) => setTimeout(resolve, wait * 1000))
continue
}
if (!response.ok) throw new Error(`${response.status} ${await response.text()}`)
const page = await response.json()
for (const tweet of page.data) console.log(tweet.created_at, tweet.full_text)
cursor = page.next_cursor
} while (cursor)GET /v1/tweets/{tweet_id} — one tweet; ?ids= for up to 100GET /v1/tweets?since=&until= — every tweet in a window of up to 31 daysGET /v1/tweets/search?q= — full-text searchGET /v1/accounts/{account} — profile by ID or username; /tweets, /likes, /followers, /following beneath itGET /v1/members — every current memberGET /v1/me — your key and its limits{data, next_cursor, request_id}. Page until next_cursor is null, not until a page is short.429 or 503, wait Retry-After seconds and retry.422 means the request was too broad for its query budget; narrow it, for example with a date range.limit.Find tweets by meaning rather than exact words. The service embeds your query with the corpus model (Qwen/Qwen3-Embedding-4B) and returns the nearest tweets by cosine similarity from the consent-filtered corpus, refreshed every ten minutes. No key is needed; each client gets 100 requests a minute. Results carry the tweet ID, so hydrate them through GET /v1/tweets?ids= above when you need counts, media, or the conversation.
curl -s -X POST https://embed.tweetstack.app/embeddings/search \
-H "Content-Type: application/json" \
-d '{
"searchTerm": "you can just do things",
"k": 10,
"threshold": 0.6,
"filter": {
"must": [
{ "key": "username", "match": { "value": "exgenesis" } },
{ "key": "created_at", "range": { "gte": "2025-01-01T00:00:00Z" } }
]
}
}'POST https://embed.tweetstack.app/embeddings/search with a JSON bodysearchTerm (text) or a 1,024-dimension vector from the same model, not bothk results (default 10, up to 1,000) above threshold similarity (default 0.65; lower it for broader results)filter with must, should, and must_not arrays of {key, match} or {key, range} conditionsGET /embeddings/stats for the current index size{success, results, count}; each result has key (the tweet ID), distance (similarity, higher is closer), and metadata with username, account ID, text, and created_at.username, account_id, or created_at for speed; a text substring filter runs after retrieval and can return fewer than k.429, wait for the X-RateLimit-Reset seconds and retry. A 400 with {success: false, error} means the body was invalid.Until 19 November 2026 the corpus tables were readable through Supabase PostgREST with a shared public anon key. That access has ended: tweets, likes, follows, mentions and the views and functions over them now refuse the anon key. Move to API v1 above with the call-by-call migration guide, or download the daily export for bulk work.
The repository contains schema guidance and complete examples.