The Autonomous Author / Page 08 — Changelog

Changelog —
every version, every decision.

Semantic versioning. Every change is tagged with type (Added, Changed, Fixed, Removed, Breaking), the requirement it satisfies from Page 02, and a brief rationale. This changelog is the living record of how the architecture evolved — not just what changed, but why.

v0.5.0 2026-03-01 Current Minor
Portfolio site
  • AddedPages 07–09 (Glossary, Changelog, Readme) completing the nine-page portfolioC-05
  • AddedScroll-reveal animations across all pages — staggered entry for diagrams and card gridsUX
  • AddedActive sidebar link tracking via IntersectionObserver on all pagesUX
  • ChangedNav link set standardised across all 9 pages — consistent topnav on every pageUX
Agent pipeline (design)
  • AddedReview Prep Agent (A-06) state machine — prioritisation logic and export_ready gate formally specifiedAR-02
  • AddedXAI Reasoning Card anatomy diagram (Diagram 16) with annotated field calloutsAR-01
  • AddedFull end-to-end sequence diagram (Diagram 17) from paste to export with timeline barBR-01
MLOps
  • AddedModel upgrade protocol with hypothetical comparison table (Llama 3.1, Llama 3.2, Mixtral)AR-06
  • Addedrules.json sample structure with R-047 and R-074 as documented examplesAR-04
  • AddedDrift signal taxonomy — three sources with detection mechanism and response protocolAR-04
v0.4.0 2026-02-14 Minor
Portfolio site
  • AddedPages 04–06 (Agent Design, MLOps, Infrastructure) with all 19 diagramsC-05
  • AddedRebuttals & Pushbacks sections on PG 04, PG 05, PG 06 — 10 total ADR-format challenges documentedAR-10
Agent design
  • AddedFormal state machine specifications for A-01 (Intake) and A-03 (Draft) — all states and transitions documentedAR-06
  • AddedDraft Agent P1 vs P2 output structure comparison (Diagram 15) — section sets, voice, temperatureBR-04
  • AddedAmbiguity Detector critique-mode prompt stance documented — "You are a critic, not an author"BR-03
  • ChangedAmbiguity Detector detection categories expanded: implicit assumptions added as fourth categoryBR-03
Infrastructure
  • AddedFull client-side reference architecture diagram (Diagram 23) — all 5 layers namedAR-03
  • AddedData flow and privacy map (Diagram 24) — three zones: always-in-browser, sent-to-Groq, never-sentAR-03
  • AddedBrowser storage schema — full field-level documentation for IndexedDB, sessionStorage, localStorageAR-05
  • AddedAES-256 encryption of Groq API key in localStorage using Web Crypto API device-derived keyAR-03
  • AddedCSP header specification — blocks all external connections except Groq APIAR-03
v0.3.0 2026-01-28 Minor
Portfolio site
  • AddedPage 03 (Design & Architecture) with 6 diagrams — DDLC-ADM phases, To-Be value streams, data model, tech architecture, gap analysisC-05
  • AddedInteractive requirements catalogue tab switcher (Business / Architecture / Constraints)UX
Architecture
  • AddedCanonical data model (Diagram 10) — 6 entities: DocumentSession, AgentRun, XAICard, DraftDocument, ComplianceReport, AmbiguityReportAR-05
  • AddedRequirements traceability matrix — all 17 requirements mapped to DDLC-ADM phases and pipeline componentsAR-10
  • AddedThree-horizon migration roadmap — H1 (foundation), H2 (workflow modes), H3 (live demo)C-05
  • Added6 ADRs documented in Rebuttals & Pushbacks on PG 03 — client-side, Groq, 5-agent split, rules.json, LangGraph-pattern, separate Ambiguity DetectorAR-10
  • ChangedArchitecture principles expanded from 4 (PG 01) to 12 (PG 03 Preliminary Phase) — P-09 through P-12 added for DDD and storage concernsAR-06
v0.2.0 2026-01-15 Minor
Portfolio site
  • AddedPage 02 (Client Brief & Requirements) — Maya Chen anchor client profile, 3 stakeholders, As-Is DDLC, MoSCoW requirements, AI readiness, 6 use casesC-05
  • AddedAs-Is DDLC detailed diagram (Diagram 06) with feedback loop arrows and latency callouts for both personasUX
  • AddedAI readiness assessment with animated bar charts across 5 dimensionsUX
Anchor client definition
  • AddedMaya Chen (Senior Technical Writer, Orbis Cloud) defined as fictitious anchor clientPortfolio
  • AddedStakeholder register — S-01 Maya, S-02 Priya Nair (Engineering Lead), S-03 James Okafor (PM)Portfolio
  • AddedPain metrics: 2.8-day SME wait, 43% violation rate, 0 DDD specs with ambiguity check, 11-day avg cyclePortfolio
  • AddedADR-001 (client-side pattern) and ADR-002 (Groq selection) documented with alternativesAR-10
v0.1.0 2026-01-01 Initial release
Portfolio site — initial
  • AddedPage 01 (Overview) — Why Now, The Problem, Philosophy (4 principles), Architecture (3-layer concept)C-05
  • AddedConvergence diagram (Diagram 01) — 3 forces plotted across 2022→2025 Q3UX
  • AddedAs-Is DDLC pain map (Diagram 02) — both personas, failure nodes, dashed absent stepsUX
  • AddedThree-layer architecture overview (Diagram 03) — Writer Interface, Agent Pipeline, Client-Side InfrastructureUX
  • AddedDesign system established — Playfair Display + JetBrains Mono + Newsreader, forest green + warm cream palettePortfolio
  • AddedRepository structure: /prompts/, /agents/, /rules.json, .github/workflows/deploy.yml, prompt-manifest.jsonAR-03
  • AddedGitHub Actions workflow skeleton — validate, hash, eval (stub), deploy jobsC-01
Design decisions (v0.1.0)
  • AddedTwo-persona model (P1 feature docs, P2 DDD specs) established as core architectural scopeBR-04
  • AddedFive-agent pipeline concept: Intake → Research → Draft → Compliance → Review PrepAR-06
  • AddedAmbiguity Detector as P2-exclusive sixth agent — separate from Draft Agent (ADR-006 precursor)BR-03
  • AddedXAI-first principle: reasoning cards designed into output schemas before any agent implementedAR-01
  • Addedrules.json as static versioned artefact rather than RAG approach — ADR-004 reasoning establishedAR-04
  • AddedHuman gate as non-bypassable UI state — export_ready flag in sessionStorage, not a UI toggleAR-02
← PG 07 — Glossary PG 09 — Readme →