SPB Git
spboucher.ai

spb/ai-atlas

Public
page 1

  1. Sat, Sep 12, 2026 2

    1. ontology: official family orgs, effort-config helpers, chained/budget effort suffixes, hyphenated family versions, property-level normalisation, run keys split from task keys
      …
      - models: FAMILY_ORGS / official_orgs / is_official_org; effort_config() and base_name() shared by the writer and the canonicalizer;
  effort suffixes chain ("-thinking-64k-high-effort"), LiveBench budgets ("32k thinking", "(no thinking)"); "max"/"fast"/"instant" are model
  tiers (Qwen3-Max, GPT-5.1-Codex-Max, Grok 4.1 Fast), no longer effort markers; family_release_hint reads "gpt-5-4" as 5.4 and never a size (38B)
- taxonomy: normalize_property(entity_type, prop, value) → (canonical, value_raw, mappings); unknown values kept, mapped to NULL
- benchmarks: RUN_KEYS (release, index_version, version, dataset_revision) leave config_key so a newer run replaces the older one; they are
  condition keys for comparability (partially comparable)
- tests/test_ontology.py: licence table, openness derivation, taxonomy, 22 model names, comparability trio, metric bounds/trust

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 4 files +350 −21
    2. Canonical upgrade foundation: ontology package (licences, openness, taxonomy, model naming, benchmarks, anomalies), migration 0003, Facts hierarchy hints, implementation audit
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 12 files +1,624 −1

  2. Fri, Sep 11, 2026 19

    1. Worker reclaims this host's running jobs at start-up (restart-safe)
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 1 file +10 −2
    2. admin requeue-dead: honour the dedupe partial unique index
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 1 file +11 −3
    3. Worker sticks to one job kind while any remain; embeddings only queued when no extraction is pending (no model thrash on llm-api)
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 2 files +13 −6
    4. LLM gateway: wait-and-retry on 503/429 while the local server loads or swaps models; one LLM job at a time in production
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 2 files +20 −4
    5. writer: normalise nested dict values before comparison
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 2 files +3 −1
    6. SDK: never lose a failed target silently, order-insensitive list comparison, unique result dedupe suffix; scheduler runs connectors sequentially; ecosystem connector fixes and registry fragments
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 7 files +40 −17
    7. arXiv: documented robots exception for the export API (arXiv API Terms of Use), 4 req/min
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 2 files +5 −1
    8. Breakage detection only on full re-extraction; first-run crawls each connector once
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 2 files +6 −7
    9. Lint fixes
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 20 files +113 −52
    10. Web wave 2: compare, timeline, price index, leaderboards, hardware fit, history/as-of/diff, graph explorer, admin console, typed listings, OG images
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 77 files +6,501 −253
    11. Connectors: 11 official labs + Hugging Face, arXiv, OpenReview, GitHub, PyPI, OpenRouter, provider pricing, leaderboards, hardware specs; SDK fixes (tree-preserving text extraction, entity unescape, resolver identifier guard)
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 121 files +115,685 −13
    12. API polish: provenance source names, methodology labels, status vocabulary, zero-filled stats
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 7 files +106 −7
    13. Web core: Next 16 app (design system, shell, home, search, models, companies, entity pages, changes, explore, methodology, sources, developers, sitemap); provider slugs -api
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 81 files +8,142 −1
    14. LLM facts: one tier below source, plausible model names only, status vocabulary normalised
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 1 file +34 −6
    15. API v1: FastAPI routers per docs/API.md (public + admin), detail builder, hardware fit, entity merge, indexes, tests
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 29 files +3,317 −4
    16. Worker: acquire the concurrency slot before claiming a job
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 1 file +8 −4
    17. Extraction schemas: tolerant coercion of LLM output (null lists, string booleans/ints)
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 1 file +73 −11
    18. Deploy: mld manifest for M2M32c, render/first-run scripts, off-node backup, DEPLOY.md; embedding model id
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 7 files +226 −2
    19. AI Atlas foundation: schema, connector SDK, archive, temporal writer, LLM gateway, scheduler, registry, CLI, Anthropic connector
      …
      Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
      Simon-Pierre Boucher · 72 files +8,918
Menu

Theme
↑↓ navigate ↵ open esc close / or ⌘K anywhere