Files
openwiki-wiki/sources/README.md
T
2026-08-22 02:17:39 +08:00

43 lines
8.9 KiB
Markdown

---
type: "Reference"
title: "Sources"
description: "Per-connector evidence indexes and ingestion coverage for the local knowledge wiki; durable synthesis lives on canonical pages."
tags: [sources, connectors, evidence, ingestion, 2026-08-21]
timestamp: 2026-08-21T18:35:00Z
---
# Sources
Per-connector evidence indexes. Each connector gets its own page (for example `gmail.md`, `slack.md`) that records ingestion coverage, freshness, and a compact evidence trail. Durable synthesis lives on the canonical pages (`/themes.md`, `/commitments.md`, `/personal-logistics.md`, `/open-questions.md`); these source pages are the evidence layer underneath.
## Active connectors
| Connector | Page | Status |
| --- | --- | --- |
| Hacker News (`hackernews-1`, LLM-agent topics) | [/sources/hackernews.md](/sources/hackernews.md) | active (last 2026-08-21 18:00 UTC — tenth pull since 2026-08-07 with an empty / noise-only Algolia lane on vendor queries, all six Algolia queries (`organoid`, `CelVivo`, `3Brain`, `Cellink`, `openwiki`, `deepagents`) returning fully-zero hits — a stricter all-zero Algolia lane than the 2026-08-20 noise-only lane, with no application-domain thematic hits; the strongest named-stack-adjacent item is `49386163` "DeepSeek-v4-flash-vision-exp" 379 pts top+best (DeepSeek vision capability extension — DeepSeek V4 Flash is the canonical DeepSeek-M3 comparison benchmark cited on the `MiniMax-m3` row via benchlm.ai 2026-08-01, V4 vision is a V4-Flash capability milestone that expands into the same multimodal territory MiniMax-M3 occupies per the MyClaw.ai 2026-07-29 confirmation), with three secondary adjacents flagged for traceability: `49390739` Show HN Proliferate 11 pts (self-hostable open-source "Codex for any coding agent" — **Codex CLI is on Robin's named stack** but Proliferate is not, so adjacent-not-promoted per scope), `49378957` "The August 17 outage" 617 pts best / 722 cmts (GitHub blog post-mortem on the August 17 incident affecting Actions + Packages + Pages + Codespaces + API + webhooks; off-scope per scope but the highest-engagement infrastructure post-mortem of the pull), `49390463` "Building an (almost) fully self-hosted, sandboxed, agentic software factory" 16 pts top (4th entry in the 5-day self-hosted-agentic-substrate cluster after DeepSeek Harness 2026-08-13 + machine0 2026-08-18 + OneCLI 2026-08-19; cluster is now a recurring-signal watchlist candidate)) |
| Git repo (`git-repo-1`, Obsidian vault + OpenWiki wiki) | [/sources/git-repo.md](/sources/git-repo.md) | active (last 2026-08-21 18:09 UTC — 0 new vault commits since 2026-08-20T18:09Z fetch (vault silent for the 5th consecutive fetch, **5 calendar days stale** since the 2026-08-16 Quintard + Guo cluster); 2 new automated wiki commits since 2026-08-20T18:09Z fetch (`b6dc337` daily update 2026-08-21 + `f6c2576` sources-sync 2026-08-21, the seventh consecutive sources-sync re-applying the 2026-08-15 count-data wipe pattern)) |
| Web Search (`web-search-1`, Tavily) | [/sources/web-search.md](/sources/web-search.md) | active (latest pull 2026-08-03; 2026-08-08, 2026-08-10, 2026-08-12, 2026-08-15, 2026-08-17, 2026-08-18, 2026-08-19, and 2026-08-20 update attempts skipped — connector not enabled, `TAVILY_API_KEY` missing) |
Additional connectors appear here automatically once they are authenticated and ingested at least once.
## Connector roll
| Connector | Status | Auth / env var | Notes |
| --- | --- | --- | --- |
| Gmail (`google`) | not configured | `OPENWIKI_GOOGLE_*` via `openwiki auth google` | Direct API; defaults to `newer_than:1d`. |
| Slack | not configured | `openwiki auth slack` | Needs `search:read` for definitive self-message search; otherwise falls back to recent conversations. |
| Notion | not configured | Notion MCP credentials | Discovery via `openwiki_list_mcp_tools`; do not guess tool names. |
| X / Twitter | not configured | `OPENWIKI_X_ACCESS_TOKEN` via `openwiki auth x` | Streams: `home_timeline`, `user_posts`, `mentions`, `bookmarks`, `list_posts`. |
| Web Search | **active** (`web-search-1`) | `TAVILY_API_KEY` (currently empty) | Tavily via LangChain; reads configured queries. Latest successful pull 2026-08-03 18:08 UTC; the 2026-08-08, 2026-08-10, 2026-08-12, 2026-08-15, 2026-08-17, 2026-08-18, 2026-08-19, 2026-08-20, and 2026-08-21 update attempts were skipped because the connector was not enabled in the runtime registry (ninth consecutive skip on the same code path; nightly searxng-fallback slot for 2026-08-20 is also empty on disk). Evidence at [/sources/web-search.md](/sources/web-search.md). |
| Hacker News | **active** (`hackernews-1`) | _(public, no auth)_ | Public feeds (`top`, `new`, `best`) + Algolia HN search queries. Source-specific query set: `organoid`, `CelVivo`, `3Brain`, `Cellink`, `openwiki`, `deepagents` (LLM-agent topics per the source-specific instruction "Track organoid / CelVivo / 3Brain / Cellink / openwiki / AI agent stories"). Last successful pull 2026-08-21 18:00 UTC — **tenth pull since 2026-08-07 with an empty / noise-only Algolia lane on vendor queries** (after the 2026-08-08 03:15 + 2026-08-08 18:00 + 2026-08-13 + 2026-08-15 + 2026-08-16 + 2026-08-18 + 2026-08-19 fully-zero pulls + the 2026-08-10 + 2026-08-12 + 2026-08-17 + 2026-08-20 noise-only pulls where vendor queries returned substring-fragment noise; the **2026-08-21 pull returns to fully-zero across all four vendor queries + both workflow queries — a stricter "all-zero Algolia lane" than the 2026-08-20 noise-only lane**, with no application-domain thematic hits to soften the absence; empty-pulls-on-vendor-queries is normal cadence, not an outage). The strongest named-stack-adjacent item of the pull is `49386163` "DeepSeek-v4-flash-vision-exp" 379 pts top+best (DeepSeek vision capability extension; DeepSeek V4 Flash is the canonical DeepSeek-M3 comparison benchmark cited on the `MiniMax-m3` row via benchlm.ai 2026-08-01 + the unsloth `unsloth/DeepSeek-V4-Flash-GGUF` quantized local-LLM run path; V4 vision is a V4-Flash capability milestone that expands into the same multimodal territory MiniMax-M3 occupies per the MyClaw.ai 2026-07-29 confirmation); three secondary adjacents flagged for traceability: `49390739` Show HN Proliferate 11 pts (self-hostable "Codex for any coding agent" — Codex CLI is on Robin's named stack so this is the strongest named-stack direct cross-walk of the pull by name, but Proliferate is not on the named stack and 11-pt engagement is below the 21-pt OneCLI / 19-pt machine0 / 46-pt MCP Memory / 100-pt vendor release bar; adjacent-not-promoted per scope), `49378957` "The August 17 outage" 617 pts best / 722 cmts (GitHub post-mortem; off-scope per scope but the highest-engagement infrastructure post-mortem of the pull), `49390463` "Building an (almost) fully self-hosted, sandboxed, agentic software factory" 16 pts top (4th entry in the 5-day self-hosted-agentic-substrate cluster after DeepSeek Harness 2026-08-13 + machine0 2026-08-18 + OneCLI 2026-08-19 — cluster is now a recurring-signal watchlist candidate). The 2026-08-20 pull (prior) produced the `class-organoid-3d-bioprinting` theme-row Last-seen refresh 2026-08-12 → 2026-08-20 via the recurrence criterion (the `49366519` Nature organoid-time-passage paper 18 pts + `49373812` Science Alert cross-walk — second neural-organoid HN signal in 9 pulls); that refresh carries forward unchanged into the 2026-08-21 pull (no organoid / 3D-bioprinting item on the 2026-08-21 pull). All four 2026-08-21 adjacents stay adjacent-not-promoted per scope discipline. Evidence at [/sources/hackernews.md](/sources/hackernews.md). |
| Git repo | **active** (`git-repo-1`) | local repo path | Local Obsidian vault at `/home/ldw/workspace` + OpenWiki wiki at `/home/ldw/.openwiki/wiki` (two configured repos per `~/.openwiki/connectors/git-repo/config.json`). Writes compact manifest; repo itself is source of truth. Evidence at [/sources/git-repo.md](/sources/git-repo.md). Last updated 2026-08-21 18:09 UTC — **0 new vault commits in the 24-hour window** (vault silent for the 5th consecutive fetch; the 2026-08-16 Quintard 2024 + Guo 2026 cluster remains the most recent vault activity, **5 calendar days stale**); 2 new automated wiki commits since the 2026-08-20 fetch (`b6dc337` daily update 2026-08-21 + `f6c2576` sources-sync 2026-08-21, the **seventh** consecutive sources-sync re-applying the 2026-08-15 count-data wipe pattern after `997b648` 2026-08-16 + `be49285` 2026-08-17 + `1fc73ae` 2026-08-18 + `dcb53a2` 2026-08-19 + `bd30339` 2026-08-20 + now `f6c2576` 2026-08-21). |
## How new source pages get created
When a connector ingests for the first time, an update run will:
1. Write the raw dump to `~/.openwiki/connectors/<connector>/raw/`.
2. Create `/sources/<connector>.md` with the ingestion manifest, freshness, and compact evidence notes.
3. Route durable items into the canonical pages and add a row in `/themes.md` if a recurring signal emerges.
Until then, the new connector only has a row in the active connectors table above and no evidence page of its own..its own..