daigest（daigest 搜索）

Daigest — Context Feed API for AI Agents

Give your AI grounded context, not hallucinated search results.

Daigest delivers AI-summarized digests from trusted sources (RSS, web pages, Slack, Notion, GitHub, and more) — grounded in real content, always traceable to the original source. Agents can also create and manage their own feeds via this API when they need context on new topics.

When AI agents search the web, they risk hallucination and unreliable results. Daigest solves this: instead of searching and hoping for accuracy, the agent subscribes to trusted sources and gets pre-summarized, factual context on demand. No guesswork, no fabrication — just reliable information the agent can act on.

Core Concepts

Authentication

Authorization: Bearer dk_live_xxxxxxxxxxxx

API keys: https://daige.st/account/api-keys

Base URL & OpenAPI Spec

https://daige.st/api/v1

OpenAPI spec for auto-generating agent tools (Claude Code, Cursor, Windsurf, ChatGPT GPTs, Dify, n8n, etc.):

GET https://daige.st/api/v1/openapi.json

Usage Flows

A. Read existing feeds

The most common flow. Feeds are already set up (by a human in the web app or previously by an agent).

GET /feeds                              → find the relevant feed
GET /feeds/{id}?since=2026-02-09T00:00Z → check for updates

If has_changes: true, use the content. If false, skip — nothing new.

B-1. Create a feed and get content via sync

POST /feeds
{
  "name": "AI News Digest",
  "memory": "Summarize the latest AI news. Highlight new model releases and benchmark results.",
  "sources": [
    { "type": "rss", "url": "https://example.com/ai-feed.xml" },
    { "type": "url", "url": "https://example.com/blog" }
  ]
}
→ 201 { "id": "feed-uuid", ... }

Option A — Sync now: Call sync immediately when the agent needs context right now.

POST /feeds/{id}/sync
→ 200 { "content": "# AI News\n\n...", "has_changes": true, ... }

Synchronous — takes 30s to 5min depending on source count. Consumes AI quota.

Option B — Schedule: Add schedule_enabled and scheduled_times to auto-sync periodically, then poll for updates.

POST /feeds
{
  "name": "Daily Tech Brief",
  "memory": "Summarize key updates from these tech blogs.",
  "schedule_enabled": true,
  "scheduled_times": [{ "time": "09:00", "days": ["mon", "tue", "wed", "thu", "fri"] }],
  "sources": [{ "type": "rss", "url": "https://example.com/feed.xml" }]
}
→ 201 { "id": "feed-uuid", ... }

GET /feeds/{id}?since=2026-02-09T09:00:00Z
→ 200 { "has_changes": true, "content": "..." }   // new digest ready

B-2. Create a feed and write content directly

The agent writes its own content to a feed — using Daigest as a context store without relying on AI sync.

POST /feeds { "name": "Research Notes" }
→ 201 { "id": "feed-uuid", ... }

PATCH /feeds/{id}
{ "content": "# Agent's research notes\n\n..." }
→ 200 { "content": "# Agent's research notes\n\n...", ... }

Useful when the agent curates and organizes information itself, and uses the feed to store or share the result.

C. Adjust an existing feed

Use PATCH /feeds/{id} to change any combination of settings in a single request.

Change sources and AI instructions, then re-sync:

PATCH /feeds/{id}
{
  "memory": "Focus on product launches only. Ignore hiring news.",
  "add_sources": [{ "type": "url", "url": "https://example.com/changelog" }],
  "remove_source_ids": ["old-source-uuid"]
}

POST /feeds/{id}/sync    → regenerate with new settings

Update content and publish as a post:

PATCH /feeds/{id}
{ "content": "# Updated digest\n\nAgent-curated content here." }

POST /feeds/{id}/posts    → publish as a post (once per 15 min)
→ 201 { "version_number": 1, "content": "...", ... }

Change schedule:

PATCH /feeds/{id}
{
  "schedule_enabled": true,
  "scheduled_times": [{ "time": "18:00", "days": ["mon", "fri"] }]
}

Endpoints

List feeds

GET /feeds
→ 200 { "feeds": [{ "id", "name", "source_count", "updated_at", ... }] }

Create a feed

POST /feeds
{
  "name": "Feed Name",                          // required, max 200 chars
  "memory": "Instructions for AI summarizer",   // optional, max 1500 chars
  "schedule_enabled": true,                     // optional
  "scheduled_times": [{ "time": "09:00", "days": ["mon", "fri"] }],  // optional
  "sources": [{ "type": "rss", "url": "..." }] // optional
}
→ 201 { "id", "name", "schedule_enabled", "source_count", "created_at" }

Get feed details

GET /feeds/{id}
GET /feeds/{id}?since=2026-02-09T00:00:00Z
→ 200 {
  "id", "name", "schedule_enabled", "scheduled_times",
  "content",      // markdown, max 2000 chars. Empty when has_changes is false.
  "memory",       // AI instructions, max 1500 chars
  "has_changes",  // false if nothing changed since `since`
  "updated_at",
  "sources": [{ "id", "type", "name" }]
}

Update a feed

PATCH /feeds/{id}
{
  "name": "New Name",                           // optional
  "content": "# Updated content",               // optional, max 2000 chars
  "memory": "New AI instructions",              // optional, max 1500 chars
  "schedule_enabled": true,                     // optional
  "scheduled_times": [{ "time": "10:00" }],     // optional
  "add_sources": [{ "type": "rss", "url": "..." }],   // optional, max 20
  "remove_source_ids": ["source-uuid"],                // optional, max 20
}
→ 200 { "id", "name", "content", "memory", "sources", "updated_at", ... }

• remove_source_ids: Get IDs from GET /feeds/{id} response sources[].id.

Delete a feed

DELETE /feeds/{id}
→ 200 { "deleted": true }

Publish a post

Snapshot the current content as a new post. Rate limited to once per 15 minutes per feed.

POST /feeds/{id}/posts
→ 201 { "version_number", "name", "content", "references", "created_at" }

Returns 429 if called within 15 minutes of the last post.

List past posts

Sync and publish both create posts. Use this to access historical digests.

GET /feeds/{id}/posts
GET /feeds/{id}/posts?cursor=42&limit=5
→ 200 {
  "posts": [{
    "version_number": 42,
    "name": "My Feed",
    "content": "# Summary...",
    "references": [{ "id": 1, "title": "Article Title", "url": "https://example.com/article" }],
    "created_at": "2026-02-08T09:00:00Z"
  }],
  "next_cursor": 38,
  "has_more": true
}

Paginate with cursor (version_number, exclusive) and limit (default 10, max 50).

Get a specific past post

GET /feeds/{id}/posts/{versionNumber}
→ 200 { "version_number", "name", "content", "references", "created_at" }

Returns 404 if the version doesn't exist.

Sync a feed

POST /feeds/{id}/sync
→ 200 { "id", "name", "content", "has_changes": true, "sources", ... }

Fetches latest data from all sources, then AI generates/updates the digest. Synchronous — waits until completion. Consumes AI quota.

Source Types

Create via API

Configure in web app, use via API

These sources require setup in the Daigest web app first, then you can read and sync feeds containing them via the API.

Best Practices

Memory examples

• Language: "Write the digest in Korean." or "Summarize in Japanese with English technical terms."
• Focus: "Focus on pricing changes and product launches. Ignore job postings."
• Format: "Use bullet points grouped by topic. Start each section with a one-line summary."
• Tone: "Write for a technical audience. Be concise — no fluff."

Sync cost management

Each sync consumes an AI request from the user's plan.

• Prefer scheduled syncs (schedule_enabled) over frequent manual syncs
• Combine related sources into one feed rather than many small feeds
• Use since to check for changes before deciding to sync

Source combination

RSS for regular updates + URL for specific pages gives comprehensive coverage. For example, an RSS feed for blog posts + a URL for the changelog page.

Rate Limits

Returns 429 with Retry-After header (seconds) when exceeded.

Error Responses

{ "error": "description", "code": "ERROR_CODE" }

Links

• Website: https://daige.st
• OpenAPI spec: https://daige.st/api/v1/openapi.json
• API keys: https://daige.st/account/api-keys
• Support: help@daige.st