Files

T

JesseMarkowitzandClaude Opus 5 480414efe0 M7: a first-class imported knowledge library

A campaign can import local .txt and .md files as Canon, Reference or
Inspiration, and the class is load-bearing rather than a label: it decides the
words a passage is framed with in the prompt, the weight it carries when
passages are ranked, and which budget it competes in when the context is tight.

This is a separate subsystem, which is the Phase 0B decision
(IMPORTED-KNOWLEDGE-DESIGN.md §73). Story Cards do not carry classification,
provenance, content identity, chunking, an index or a lifecycle, and they were
not promoted into something that does. Nothing here reads or writes one.

The subsystem, in backend/app/knowledge/:

  classes      the three classes, their weights, and the prompt framing
  chunking     deterministic, heading-aware, 60-800 tokens, no overlap
  fts          SQLite FTS5 with porter stemming; scoped and bounded in SQL
  importer     validate, hash, store, chunk, index — in one transaction
  embeddings   local Ollama vectors through the shared provider
  retrieval    query construction, hybrid merge, rerank
  inject       the budgeted cut and the rendered prompt sections

Relevance admission is a separate stage from ranking, and that separation is
the milestone's most expensive lesson. An independent review found the first
implementation deciding relevance with a floor expressed as a share of the best
candidate — which the best clears by construction — so a passage was admitted on
every turn regardless of the scene. A query about tide tables and container
tonnage retrieved all five sources of a fantasy campaign, narrator-only hidden
Canon among them.

So the pipeline is now:

  candidate generation -> admission -> ranking -> class weighting -> budget

Admission reads raw, candidate-set-independent signals: the cosine the model
returned, and how many distinct meaningful query terms a passage contains.
Ranking reads normalized ones, because bm25 has no fixed range and cosine's zero
is not zero. Normalization decides order among things that matched; it can never
decide whether anything matched. Authority is applied after admission, so a
class orders what matched and never rescues what did not.

Retrieval may therefore return nothing, and on a scene unrelated to the library
it does.

The other decisions that each replaced an obvious wrong one:

- The class multiplies relevance rather than adding to it. An additive bonus
  satisfies "Canon outranks Reference" and makes "do not include irrelevant
  Canon" impossible, because a large enough constant wins on its own.
- The semantic floor is measured, not guessed: 113 production-path pairs against
  nomic-embed-text put targeted matches at 0.55-0.85 and off-topic pairs at
  0.36-0.56, and 0.58 sits between them. Because it is a property of that model
  and not of cosine similarity, it is keyed to the model rather than applied to
  whatever is configured: an embedding model with no measured calibration in
  this build does not borrow the number. Semantic admission is skipped, the
  campaign retrieves lexically, and the reason is stated in the knowledge status
  and in the turn's provenance. Degrading to lexical keeps the library usable;
  lending the threshold to an unmeasured model is how the admitted-everything
  defect would return.
- One lexical term is not evidence. Two distinct meaningful terms, or one that
  is neither a standing campaign entity nor a negligible share of the query.
  The stop list grew from 42 words to 261, all function words — no subject
  matter, because a stop list that removes subject matter stops finding "The
  Silver Key".
- Lexical retrieval is a production path, not a fallback. It finds the proper
  nouns and invented terms a setting bible is made of, and the library is fully
  usable with no embedding model configured.

Safety is structural rather than filtered. Imported text reaches the prompt
whole, inside a section that says what it is, under a rule stating the authority
order in words and refusing every instruction inside it. No endpoint accepts a
filesystem path, so H08 has no mechanism to escape from. Nothing renders
imported content as HTML, so a script tag is five visible characters and a
remote image is never fetched. Import, chunking, indexing, retrieval and a turn
open no socket at all; only embeddings do, through the endpoint allowlist the
memory bank already uses.

Provenance is the rendered text, not a foreign key: deleting a source cannot
turn a historical turn's evidence into dangling ids.

Schema: knowledge_sources, knowledge_chunks, knowledge_embeddings, and an FTS5
virtual table attached to knowledge_chunks as a DDL hook so it is created and
dropped with the table it indexes. Migration 92. A pre-M7 database opens
unchanged and needs no sources to play.

Bundle: the source content and the reader's judgements about it travel; the
passages, index rows and vectors are rebuilt on import, so a restored campaign
is searchable immediately without a reindex step.

One runtime dependency: python-multipart, Starlette's multipart parser. It is
what makes the upload surface possible, and the upload surface is why no
pathname is ever accepted.

The test doubles were the reason the defect shipped, so they were corrected too.
The retrieval stub scored unrelated text at 0.06-0.20 where the real model
scores it at 0.43-0.44, and its docstring said it had deliberately removed the
constant component that "would put a similarity floor under every pair" — which
is exactly the property real models have. The stub now has that floor, one test
fails if it is ever removed, and another reproduces the superseded rule and
asserts it is still fooled by the same fixture. Run against the pre-corrective
implementation, the new suite fails 13 of 18.

Tests: 939 passed, 14 skipped (836/7 at M6). 110 new across seven files, one of
which mocks nothing between itself and Ollama and re-measures the similarity
separation on every run. 43/43 checks in a real Firefox, reproduced.
Docker build clean.

Four other defects found by review or by the browser run were fixed here rather
than carried: an unreachable relevance constant that appeared to enforce
something and did not; acceptance tests using the wrong fixture files, so G07's
trap was never exercised; a bidirectional override surviving into displayed
filenames; and, from the implementation pass, the Insights panel showing M5's
two state sections as raw keys and the source inspector refetching on every
keystroke.

M7 was independently reviewed, which returned PASS WITH CORRECTIVE WORK
REQUIRED. Both blocking findings are closed, and closeout resolved the
embedding-model calibration boundary the corrective pass had left as debt.
planning/reports/M7-IMPLEMENTATION-REPORT.md carries the review, the corrective
closeout and the closeout verification in sequence, none overwriting another.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017HdaXiFbscatQaLS7dJk6b

2026-09-06 15:40:13 -04:00

26 KiB

Raw Permalink Blame History

Planning Package Version

Package: Adventure Storyteller Planning Package v3.0
Revision date: 2026-09-06
Status: Phase 0 complete; architecture selected; Milestones M1-M7 implemented and accepted; M8 is next to brief.

v3.0 — M7 Closeout (2026-09-06)

M7 is complete. Two items the corrective pass had left open are resolved.

The embedding-model calibration boundary. SEMANTIC_FLOOR = 0.58 was measured against nomic-embed-text, and the corrective pass documented only the safe half of that: a model scoring everything lower degrades to lexical-only. A model scoring unrelated material higher would have recreated M7-F1 on a build whose tests all pass. Semantic admission is now per model: an uncalibrated model does not inherit the threshold, semantic retrieval is skipped for it with the reason reported, and the library degrades to lexical-only. Recorded in TECHNICAL-DESIGN.md §13.3 and IMPORTED-KNOWLEDGE-DESIGN.md §76.

The ambiguous export/import 53/54. The 54th case was a false positive in the independent review's own harness — its "no filesystem path" assertion was a substring test that fired on text/markdown, a MIME type. Replaced with three precise checks; the suite is 56/56 and no product behaviour was involved.

What closeout changed in the active documents:

TECHNICAL-DESIGN.md §13.3 — new. A similarity threshold is a property of the model, the two ways a different model breaks it are not symmetric, and the product refuses to apply a threshold to a model it has not measured.
IMPORTED-KNOWLEDGE-DESIGN.md §76 — the same, in the design's own terms: §25's local-Ollama embedding stands; what is added is that a threshold must be measured before it is trusted.
BUILD-MILESTONES.md M7 — marked COMPLETE, with the capabilities later milestones inherit and the debt carried forward, including that calibrating further embedding models is a measurement rather than a guess.
V1-ACCEPTANCE-TESTS.md — the M7 results promoted from implementation-pass evidence to reviewed results.

v2.9 — M7 Independent Review and Corrective Pass (2026-09-06)

The review returned PASS WITH CORRECTIVE WORK REQUIRED. It closed the five acceptance conditions the implementation had flagged as unmeasured — C05, G06, G07, G10 and hidden Canon, all exercised against a real narrator and all passing — and found two blocking defects, both now corrected.

M7-F1 — imported knowledge was injected regardless of relevance. Relevance was decided by a floor expressed as a share of the best candidate, which the best clears by construction. A query about tide tables and container tonnage retrieved all five sources of a fantasy campaign, narrator-only hidden Canon among them. Corrected by separating relevance admission from ranking.

M7-F2 — the retrieval suite could not detect it. Its stub scored unrelated text an order of magnitude lower than the real model, so the broken gate passed. Corrected with a stub that has the real model's similarity floor, plus a test that fails if the floor is removed and one that shows the superseded rule still being fooled. The new suite fails 13/18 against the pre-corrective code.

What the corrective pass forced into the active documents:

TECHNICAL-DESIGN.md §13.2 — new. Relevance admission is a separate stage from ranking, and the general rule behind it: a relevance decision must rest on a signal meaningful on its own, because normalization answers "which of these is best" and can never answer "is any of these any good". A pipeline that ranks first and cuts second has no way to return nothing.
TECHNICAL-DESIGN.md §13.1 — the pipeline diagram gains the admission stage.
IMPORTED-KNOWLEDGE-DESIGN.md §76 — retrieval corrected: admission before authority, and the plain statement that retrieval may return nothing, which is what §30 means when every source is irrelevant.
BUILD-MILESTONES.md M7 § Status — both findings, their corrections, and the model-specific calibration recorded as carried debt.

Three non-blocking findings were folded in: a relevance constant that could never fire was removed rather than re-tuned; the acceptance tests moved to the standard TEST-CAMPAIGN-FIXTURE.md §12 files so G07's trap is finally exercised; and Unicode format characters are stripped from displayed filenames. A pre-existing M5 narrator-protocol issue was recorded and deliberately left with M5.

v2.8 — M7 Implementation Pass (2026-09-06)

Not a closeout. M7 is implemented, not accepted, and this revision records what the implementation pass built and measured so that an independent review has something to verify against. No milestone report was written: the convention this package follows puts the report in reports/ and has the reviewer write it, treating the build summary as claims to check.

M7 — First-Class Imported Knowledge Library. A campaign can import local .txt and .md files as Canon, Reference or Inspiration; retrieval is hybrid (SQLite FTS5 plus local Ollama embeddings), reranked by relevance × class, bounded by its own token budget, framed in the prompt as untrusted data with the authority order stated in words, and fully traceable in the Insights panel. It is a separate subsystem: AI-DnD's Story Cards were not promoted into it and are untouched.

What implementation forced into the active documents:

TECHNICAL-DESIGN.md §13.1 — new. The implemented pipeline, and four decisions that each replaced an obvious wrong one: the class multiplies relevance rather than adding to it; both retrieval scores are normalized per query against the best of their own path; the relevance floor is therefore relative rather than absolute; and lexical retrieval is a production path rather than a fallback.
DATA-MODEL.md §24A — new. The three tables and the FTS5 virtual table, and the line between what the reader gave the campaign and what the machine derived from it. Only a source's content and its classification are not derivable.
DATA-MODEL.md §25 — the retrieval record is not a table. It lives in the turn's own context snapshot and carries the rendered text, because a table of foreign keys would turn every historical turn's evidence into dangling references the moment a source were deleted.
DATA-MODEL.md §29 — what the bundle carries for imported knowledge, and why passages, index rows and vectors are rebuilt rather than exported.
CONTEXT-AND-MEMORY.md §29, §41-42, §46 — the knowledge budget as implemented (a protected cap for always-included Canon, a share of the rest filled in authority order), always-include as a Canon-only mechanism, and hidden Canon as prompt discipline rather than as filtering.
IMPORTED-KNOWLEDGE-DESIGN.md §76 — new. Where this document offered options, which was chosen and why; and, named rather than left to be discovered, the six things it contemplates that M7 does not implement — entity linking, tags, manual priority, scene pinning, Canon-versus-Canon conflict detection, and source versioning.
V1-ACCEPTANCE-TESTS.md — results for G01-G10, C05, F05, F06, I05 and H06-H09, marked as implementation-pass evidence rather than review findings. F05 and F06 move from PARTIAL / PASS for story memory to complete. H09 is recorded NOT APPLICABLE with a test that fails if an archive extractor is ever added to this surface. C05 is recorded as a pass on the assembled prompt with the gap stated: no real narrator generation was run against it.
BUILD-MILESTONES.md M7 § Status — new. What was built beyond the scope list, and the debt carried forward, deliberately.

One runtime dependency was added: python-multipart, Starlette's multipart parser. It is what makes the upload surface possible, and the upload surface is why no endpoint in the knowledge API accepts a filesystem path.

v2.7 — M5 and M6 Closeout (2026-09-06)

Two milestones, and one lesson they share: an independent review found a real defect in each after the implementation reported success, and in both cases the defect was invisible to the tests the implementation had written for itself.

M5 — Genre-Neutral Authoritative Narrative State. Accepted 2026-09-04. Its review returned PASS WITH CORRECTIVE WORK REQUIRED: editing a narrator turn rewound the campaign's live state while the head stayed at the tip, breaking the transcript position == head == authoritative state invariant. The corrective pass rebuilt narrator editing on the §§14-15 fork semantics. Recorded here late: M5's closeout did not add a package revision entry, and this one covers it.

M6 — Branch-Safe Context, Summaries, and Long-Term Story Memory. Accepted 2026-09-06. Sequence:

Implementation. Summaries moved onto lineage-anchored rows; memory authority, provenance, an output reserve, and observable derived-work failure were added. Reported as passing, including E03.
Independent review — E03 still failed. Summary rows were anchored, but generation was seeded from adventures.story_summary, a campaign-global column with no lineage. After a divergence the summariser was handed the abandoned line's prose and asked to update it, so the new summary carried abandoned content inside a correctly anchored row. The review also found F02 passing by a single retrieval slot, two test defects, and a misleading derived-work status.
Corrective pass. Generation is seeded from summaries.current; redundancy suppression was added before the retrieval cut; the test fixture no longer makes real network calls; the browser suite gained a scenario that regenerates a summary after divergence.

What the planning package learned from M6, recorded in the active documents:

CONTEXT-AND-MEMORY.md §11 — lineage safety has two halves. Anchoring the output row is not enough; the input to the summarizer must be scoped by the same rule.
V1-ACCEPTANCE-TESTS.md E03 — a valid test must regenerate a summary after diverging. Checking only that the old row went ineligible passes while the defect is live.
CONTEXT-AND-MEMORY.md §20/§22 — what M6 implements (redundancy suppression between memories) is now distinguished from what it does not (importance, entity, recency and thread ranking; cross-layer duplication).

v2.6 — M4 Closeout (2026-09-03)

M4 is accepted. Its review returned PASS WITH CORRECTIVE WORK REQUIRED; the corrective work is done, and the browser condition that M3 and M4 both carried is closed.

The three review findings, fixed:

B-1 — the Save Point list was an N+1. It resolved each Save Point with its own query and loaded whole Action rows, narration included, to answer "does a row exist here". It is now one bulk two-column coordinate query plus one lineage computation: 53 SELECTs for 25 Save Points became 5, and the count no longer grows with the list. Guarded by three tests, including one proving the coordinate is matched as a pair — an IN-list version would report a Save Point resolved because another branch has a live row at the same depth.
B-2 — deleting a branch silently deleted its Save Points. Fixed as a behaviour defect rather than a missing warning, because STORY-BRANCH-SEMANTICS.md §19 says a named checkpoint remains until explicitly deleted and §28 already required future cleanup to retain checkpoint-referenced paths. A branch a Save Point names can no longer be deleted. The request is refused with the offending Save Points named; the user deletes them explicitly, which deletes no story, and the branch then goes. Both delete controls disable and explain. Recorded as a new §19.1.
B-3 — the D11/L03 automation never left one process. A new module spawns real server processes, kills the first, and reads the campaign back with the second.

Also fixed (review §S C-5): creating a Save Point now takes the campaign's turn lock, so "save where I am" cannot read a head a turn in flight is about to move. Rename and Delete deliberately do not take it, and a test pins that decision.

Real-browser verification — the first in this project. A Firefox 154.0.1 driven through geckodriver over the W3C WebDriver protocol exercised the rendered DOM for both milestones: 44/44 checks passed, no console errors. It covered M3's Undo/Redo enable states, transcript movement, Retry and the take pager, and divergence retiring Redo; and M4's whole Save Point lifecycle including both confirmations and the new branch-delete warning. The outstanding M3 browser condition is therefore closed as well. No dependency was added to the repository: the WebDriver client for the run was written against stdlib HTTP.

Also corrected, found while fixing B-2: models.py, TECHNICAL-DESIGN.md §8.8 and DATA-MODEL.md §8 all described the cascade as the durability rule. They now describe the refusal, and record that checkpoints.branch_id's cascade survives as referential integrity that the application no longer reaches.

Documents corrected by this closeout:

V1-ACCEPTANCE-TESTS.md records results for D11-D14, I04, L03 and the E-series, and states that the browser-level condition is satisfied. No pass condition was weakened — and D11/L03 now note that the automation crosses a genuine OS process boundary, which is the standard later milestones should meet.
DATA-MODEL.md §8 records the coordinate as implemented, with the retry measurement that settles coordinate-versus-turn-id, and three decisions that were previously implicit: names are not unique, several Save Points may name one position, and the list is newest-created first.
STORY-BRANCH-SEMANTICS.md gains §19.1 — a checkpoint protects the history it names. This is the only behavioural specification change in the closeout, and it strengthens §19 rather than weakening anything.
BROWSER-UX-SPEC.md §25 rules for the implemented vocabulary: Moment N, not Turn N, because the branch panel and tree overlay already count in moments. Vocabulary only; no behaviour changes.
BUILD-MILESTONES.md marks M4 COMPLETE, records the fixes and the browser result, and warns M5 that the instrumentation to move is now 55 tests.
README.md records M1-M4 accepted and M5 as next to brief.
PROJECT-SOURCES.md and project-sources.txt point at the current report; project-sources.txt still named the archived M3 report and was corrected.

Report rotation happened in the reporting pass that preceded this closeout: M3-IMPLEMENTATION-REPORT.md moved to archive/milestone-reports/ as a pure rename, and reports/ now holds M4's report, whose §W is this closeout's evidence record.

No new ADR. ADR 012 already decides the architecture, and the corrective work forced no new architectural decision. SPECIFICATION.md, STORY-BRANCH-SEMANTICS.md, SECURITY-THREAT-MODEL.md, CONTEXT-AND-MEMORY.md, IMPORTED-KNOWLEDGE-DESIGN.md and ADRs 003, 005 and 012 are unchanged.

M5 readiness: ready. Save Points store no state and no checkpoint code reads any, so M5 can change what a snapshot contains without touching what a Save Point is — provided it keeps state recoverable at a position without replay (TECHNICAL-DESIGN.md §10.4).

v2.5 — M4 Implementation (2026-09-03)

M4 added durable named Save Points. This revision records only what the implementation established as fact; no product requirement changed, and the milestone is not marked accepted — its review has not been written.

TECHNICAL-DESIGN.md gains §8.8 and §9.2: the Save Point as a name plus a coordinate holding no story, restore as head movement with a bounds check, the rule that the branch half of the head moves only when the coordinate is off the path being read, restore never forking, and the bundle carrying Save Points independently of the head.
DATA-MODEL.md §8 records the pointer as implemented — (branch, depth) rather than a turn id, with the reason: one coordinate holds every attempt at a turn and exactly one is live, so a coordinate follows a retry where a row id would pin a superseded take. §29 records the checkpoints in the export.
BUILD-MILESTONES.md M4 gains a status block: what shipped, the one architectural decision the milestone had to make and why it needed no new ADR, the test count, and the outstanding browser condition.
README.md describes Save Points as a user-facing capability.

No ADR was created. ADR 012 already decides the architecture M4 needed — restore reuses active-head movement — and a table is not a decision. The one question ADR 012 does not answer, whether restore moves the branch half of the head, is that same mechanism applied to a coordinate ADR 012 already defines; TECHNICAL-DESIGN.md §8.8 records the answer rather than a new ADR asserting it.

SPECIFICATION.md, SECURITY-THREAT-MODEL.md, STORY-BRANCH-SEMANTICS.md and V1-ACCEPTANCE-TESTS.md are unchanged. M4 altered no product requirement, added no outbound path, and implemented the checkpoint semantics STORY-BRANCH-SEMANTICS.md §18-25 already specified rather than amending them.

The browser smoke test remains unperformed, now for both M3 and M4. No session has had a usable browser. See BUILD-MILESTONES.md M3 and M4.

v2.4 — Documentation Consolidation (2026-09-03)

No product requirement, architecture decision or milestone status changed in this revision. It reorganises the documentation so that a new coding agent can tell authoritative material from evidence at a glance.

planning/archive/ is created and is non-authoritative by declaration (archive/README.md). It holds phase0/ — the research that chose AI-DnD — milestone-reports/ — the completed M1 and M2 reports — and decisions/, which now holds ADR 008, the Phase-0-before-build process gate that Phase 0 satisfied. ADR numbering continues from 012; 008 is not reused.
planning/reports/ now holds only the current milestone's report, M3-IMPLEMENTATION-REPORT.md, because M4 planning has to consult it. It moves to the archive when M4's report replaces it.
The Phase 0B execution prompts and handoff/status/summary documents (CODEX-HANDOFF-NOTE.md, PHASE-0B-CODEX-BRIEF.md, PHASE-0B-CODEX-HANDOFF.md, PLANNING-UPDATE-SUMMARY.md) and the Phase 0A discovery and triage reports were deleted: intermediate working documents whose conclusions all reached the two recommendation reports, and which remain in Git history.
Upstream AI-DnD's inherited plan/ build log and docs/ project site (guides, generated HTML, screenshots) were deleted. They documented a hosted, scripted, multi-user product with accounts — every screenshot showed a Scripts tab and a Sign up button — which M2 removed. Both trees remain in Git history and in upstream.
README.md, DEVELOPMENT.md and PROVENANCE.md are corrected where they pointed at the removed trees or described removed capability as present. DEVELOPMENT.md's "things M1 did not touch" section had gone stale at M2 and now says what is actually still inherited.
planning/README.md is rewritten as the documentation index: current milestone, the three-way active/ADR/archive split, the authority order, reading order, where reports live, and what comes next. The milestone correction tables are preserved unchanged.
New: planning/PROJECT-SOURCES.md and planning/project-sources.txt, the manifest of files to upload as ChatGPT Project Sources.

v2.3 — Post-M3 Closeout (2026-09-03)

M3 replaced destructive Undo with a stored active head. Its review is archive/milestone-reports/M3-IMPLEMENTATION-REPORT.md, which is also M3's primary evidence record — no separate baseline report was produced — and whose §W records this closeout.

In summary:

the architecture is recorded as ADR 012 — Active-Head Non-Destructive History: the head is stored rather than derived, every read of the story is capped at it in one place, one mechanism moves it, state comes from the node rather than from a replay, the first write below a moved-back head is the divergence, and Redo is decided by the lineage rather than by a flag. ADR 005 is unchanged: it states the product requirement, and ADR 012 states the architecture chosen to implement it,
two history semantics are ratified in STORY-BRANCH-SEMANTICS.md: Undo crosses fork points to the campaign opening (§5), and the system refuses to switch which take is live while a later story is off screen (§10),
a new §14A records the interim in-place-editing rule — refuse when story descends from the turn and is not on screen — and states explicitly that §14-15's full narrator-edit requirement stands and is completed in M5,
TECHNICAL-DESIGN.md gains §8.7 and §9.1 recording the implemented model and bundle behaviour as fact, and a constraint on §10.4: the snapshot half of the hybrid state model is a requirement, because head movement must not become proportional to campaign length,
DATA-MODEL.md records the head as campaign-stored, the branch disposition as implemented and deliberately advisory, and the export as carrying a chosen position rather than a derived one,
BUILD-MILESTONES.md marks M3 complete, states the one outstanding condition (the browser smoke test), tells M4 to reuse M3's head movement rather than build a second restore path, and gives M5 three constraints,
V1-ACCEPTANCE-TESTS.md records D03's full pass, states D10's milestone ownership without weakening any pass condition, adds I07's pre-M3 bundle clause, and resolves the apparent L01/A05 conflict,
README.md is corrected to describe the current local-only single-user application rather than the upstream hosted one.

SPECIFICATION.md and SECURITY-THREAT-MODEL.md are unchanged: M3 altered no product requirement and touched no path in the threat model.

v2.2 — Post-M2 Closeout (2026-09-03)

M2 removed the hosted, cloud, account and scripting surface and added the inference endpoint policy. Its review recommended six planning changes and reported rather than applied them; all six are applied in this revision, listed in README.md § Post-M2 corrections applied, with the evidence in archive/milestone-reports/M2-BASELINE-REPORT.md and archive/milestone-reports/M2-IMPLEMENTATION-REPORT.md.

In summary:

the inference endpoint policy is recorded as implemented — an address allowlist of explicit local-network CIDRs, every resolved address checked, enforced when settings are saved and again before every outbound request, with TLS verification never traded against it (new ADR 011, SECURITY-THREAT-MODEL.md §10A),
its two residual limits are stated rather than mitigated: a hostile host already on the trusted LAN, and the DNS-rebinding interval between the policy's resolution and the client's connection,
TECHNICAL-DESIGN.md §5.1 items 3 and 4 are resolved, and a new §5.2 records the M1/M2 production architecture as fact,
a wiring rule is added (§18.1): removing a setting requires testing a real consumer path, and adding one requires proving it reaches its component — M2 shipped two defects behind a 604-test green suite because the tests at that boundary were mocks,
BUILD-MILESTONES.md records M2 complete, warns M5 that eight rollback tests use the world-state engine as instrumentation rather than as architecture, and requires M6 to make background memory failure observable,
the security acceptance contract is strengthened: H10 now names the wildcard-origin and /api 404 conditions, and new H12 covers inference endpoint enforcement including the database-edited-behind-the-API case.

SPECIFICATION.md is unchanged: M2 altered no product requirement. Nothing in the architecture selected in v2 was reversed.

v2.1 — Post-M1 Corrections (2026-09-02)

M1 implementation evidence contradicted or under-specified parts of v2. The corrections are recorded in the documents themselves and listed in README.md § Post-M1 corrections applied; the evidence behind them is in archive/milestone-reports/M1-BASELINE-REPORT.md and archive/milestone-reports/M1-IMPLEMENTATION-REPORT.md.

In summary:

a trusted-LAN Ollama may be HTTPS with a privately issued certificate; clients verify against the operating system's CA store, with full certificate and hostname verification and no bypass option (ADR 002, TECHNICAL-DESIGN.md §5, A06),
offline claims require a fresh cache and no route out to be evidence at all, and vendored runtime artifacts should be integrity-verifiable (ADR 004),
A05's invariant is about accepted history; the user's submitted text is deliberately retained on a failed turn,
A06 requires a real second machine and an HTTPS endpoint; a plain-HTTP LAN test is no longer sufficient evidence,
the standard test environment records CPU/GPU/RAM, because cold model load on a CPU-only host exceeded the inherited 120 s client timeout,
BUILD-MILESTONES.md records M1 as complete and reframes M2's endpoint work as narrowing an existing capability rather than inventing it,
SECURITY-THREAT-MODEL.md §53 distinguishes inbound TLS (still deferred) from outbound certificate verification (required, done in M1).

Nothing in the architecture selected in v2 was reversed.

v2 — Post Phase 0B Revision (2026-09-01)

This v2 package supersedes the earlier planning package produced before the final Phase 0B review and the trusted-LAN Ollama deployment clarification.

Key v2 changes include:

AI-DnD selected as the production base at the pinned Phase 0B commit.
Non-destructive head-cursor Undo/Redo design selected.
Explicit typed/absolute narrative-state events selected for production state handling.
The authoritative narrative-state document — its shape, its authority/provenance fields, and the event/document/snapshot split — recorded in ADR 013 after M5 implemented it.
Imported knowledge separated from AI-DnD Story Cards.
Trusted-LAN Ollama inference supported in v1 while the storyteller UI/API remains loopback-bound by default.
Offline first-use dependencies and runtime remote assets identified as M1 hardening work.
Production implementation divided into milestones M1-M11.

Historical Phase 0 prompts/reports are retained as evidence and should not be treated as current implementation instructions unless a current milestone prompt explicitly refers to them.

26 KiB Raw Permalink Blame History Unescape Escape

Planning Package Version

v3.0 — M7 Closeout (2026-09-06)

v2.9 — M7 Independent Review and Corrective Pass (2026-09-06)

v2.8 — M7 Implementation Pass (2026-09-06)

v2.7 — M5 and M6 Closeout (2026-09-06)

v2.6 — M4 Closeout (2026-09-03)

v2.5 — M4 Implementation (2026-09-03)

v2.4 — Documentation Consolidation (2026-09-03)

v2.3 — Post-M3 Closeout (2026-09-03)

v2.2 — Post-M2 Closeout (2026-09-03)

v2.1 — Post-M1 Corrections (2026-09-02)

v2 — Post Phase 0B Revision (2026-09-01)

26 KiB

Raw Permalink Blame History