Undo deleted. It removed the trailing AI action and the player action in
front of it, pruned the memories covering them, and let the tip fall back
to whatever survived. That made it the one operation in the application
that destroyed accepted story, and it was why there was no Redo: the
turns to move forward into no longer existed. Phase 0B demonstrated the
head-cursor alternative in a disposable spike; this is that concept as
production code.
backend/app/head.py is the whole of it. Three questions that used to be
one — where the story is being read, how far it is retained, and where it
opens — are now three functions, and every caller that moves the head or
asks about it goes through this module. The spike put the fork check in
the write path and left Retry and Add-take on the old one; sharing the
rules is what stops that divergence coming back.
lineage.Path now caps every entry at the head, so hiding the retained
future costs nothing at the call sites: the transcript, the context
builder, attempts.preceding and memory retrieval already funnelled
through path_of and narrow together. Path.uncapped() is the deliberate
exception, and only Redo and the fork check may use it. The memory bank
needs no pruning for the same reason — a memory carries the coordinate of
the node its block ends on, so one derived past the head falls outside
the capped clause and becomes retrievable again on Redo without having
been deleted and re-embedded.
Undo alone does not fork. Moving the head is not a decision to abandon
anything, since the user may be reading or about to Redo; the first write
below the head is where the story states which continuation it means. A
head already at the tip forks nothing, so a story that is never undone
forks exactly as often as it did before and the branch table does not
fill up with one branch per turn. Redo follows the lineage rather than
choosing among branches, which is what invalidates it after a divergence
with no flag to set or clear.
Migrations 78 and 79 give a branch superseded_at and superseded_depth.
Nothing reads them to decide behaviour — Redo is decided by the lineage,
so a stale or hand-edited value here cannot make the story wrong. They
exist so the cleanup and discarded-history features left to a later
version have something to select on, and so a divergence is observable in
a test.
Deleting an action no longer drags a moved-back head forward to the
recomputed tip, which would have silently redone the story. can_undo and
can_redo ride on AdventureOut and ActionPage because the client can work
out neither for itself: the campaign opening may be off the top of the
loaded window, and the retained future is never sent to it.
This is a checkpoint, not the finished milestone. 601 backend tests pass.
Five still assert the destructive contract — they count rows after an
undo and expect the story to be shorter — and need rewriting against the
new one; the world-state assertions inside them already pass. Export and
import do not yet carry the head coordinate, so a bundle still reopens at
the deepest node and can silently redo an undone story, which is the
Phase 0B finding this milestone exists to close. The browser has no Redo
control yet. None of the M3 acceptance coverage (D01-D10, E01-E04,
I01-I03, I07, L01-L02) is written.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QF5TcoB86QADgjHz1GZe8u
M2's review reported six planning recommendations rather than applying them,
three marked before M3. All six are applied here, plus three additions drawn
from the same evidence. No implementation file is touched.
The endpoint policy was the gap that mattered. It is the most consequential
setting in the application — the storyteller sends the player's prose, the
context, the memories and the embedding inputs to whatever address it names —
and it existed only as a module docstring. It is now ADR 011 and a new §10A in
the threat model, which also retires the assumption in §71A that the inherited
guard was a starting point. It was not: AI-DnD's SSRF guard blocked private
addresses to stop a hosted server reaching its own internal network, which is
the exact opposite of what a local storyteller needs. It was removed, not
adapted.
Both documents state the rule as implemented — an allowlist of explicit
local-network CIDRs, every resolved address checked, enforced on save and again
before every outbound request, TLS never traded against it — and both state the
two residual limits plainly rather than implying they are covered: a hostile
host already on the trusted LAN is inside the permitted boundary, and a
rebinding interval exists between the policy's resolution and the client's
connection. Accepted risks, not M3 work.
The CIDRs are spelled out rather than derived from is_private/is_reserved, and
the ADR records why: is_private is true of the documentation ranges and
0.0.0.0/8, and is_reserved is true of IPv6 loopback, so a rule built on it
refuses an ordinary same-host Ollama on [::1].
TECHNICAL-DESIGN §5.1 items 3 and 4 are marked done, closing all five hardening
items. A new §5.2 records the M1/M2 architecture as fact rather than intention,
so later milestones inherit what the code does. A new §18.1 carries the lesson
of M2's two regressions: when removing a setting, test a real consumer
construction path; when adding one, prove it reaches the component that uses
it. Both defects hid behind a green suite because the tests at that boundary
were mocks.
BUILD-MILESTONES records M2 complete, with the capabilities later milestones
inherit and the debt carried forward. Two notes go to milestones that would
otherwise misread what M2 left them. M5 is told that eight rollback tests now
use the world-state engine as instrumentation and not as endorsement — the
instrumentation moves when the protocol does, and those tests are reworked
rather than deleted. M6 is told that the memory bank died silently under a
green suite, so background failure must be observable and at least one real
provider-construction path must be tested.
The security contract gains what M2 demonstrated. H10 now names the two
conditions that were defects during M2: a wildcard origin must be refused at
startup, and an unknown /api path must 404 rather than returning the SPA with
200. New H12 covers endpoint enforcement, and its fourth pass condition is the
one that matters — a public endpoint written into the database behind the
settings API must still be refused at the wire. A build passing the first three
and failing that one has configuration validation only.
SPECIFICATION.md is deliberately unchanged. M2 altered no product requirement;
it removed capability the specification never asked for.
The two M2 reports gain appended closeout notes rather than edits. Their
original wording about an uncommitted working tree was true when written, and
the note records what happened afterwards: the six-file correction is 8652fe7,
8c65ae9 remains the implementation commit, and the two were never squashed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HsZBU8sWRuYTyLgWsu2oQ6
The post-implementation review of M2, plus the three corrections it took
to make the evidence true. Reports:
planning/reports/M2-BASELINE-REPORT.md 868 lines, the measurements
planning/reports/M2-IMPLEMENTATION-REPORT.md 758 lines, the reading of them
Verdict is PASS, accept with non-blocking debt, proceed to M3. Every M2
requirement is met and the ones that matter were tested by running the
build rather than reading it: a cloud endpoint written straight into
SQLite with sqlite3, behind the API's back, still refused at the wire;
trusted-LAN HTTPS against the real second machine with verification on;
captures showing zero packets outside loopback and the approved host.
Three defects, all found by running the shipped image.
The memory bank was dead. M2 removed Settings.api_key_plain with the API
key, and memorybank's two provider factories still read it. It failed
inside a fire-and-forget task, so no user error, no log anyone would
read, and no test — every memory test stubs those factories. All 604
tests passed with summaries and embeddings silently not happening.
The configurable model timeout never reached the turn engine. Stored,
validated, exposed in the API, rendered in the UI, and not passed to the
provider. M2's own exit criterion was half met: the constant had moved
but the setting did nothing.
And requirements.lock still pinned quickjs, psycopg and cryptography, so
the setup path DEVELOPMENT.md gives a new developer would have
reinstalled all three.
Both code defects now have the test that would have caught them: one
constructs every provider factory from a real Settings row, one drives
the turn endpoint, the chat endpoint and the summariser and asserts the
configured timeout arrives at each. That is the lesson worth keeping from
this milestone — after removing an attribute, build each consumer from a
real object; after adding a setting, prove it lands. Both failures were
in background or plumbing paths, which is exactly where a subtractive
change cannot see itself.
606 tests pass, up from 604. Lint, build and image are clean. Every
runtime result in the baseline report came from an image built after
these fixes; the reports say plainly that commit 8c65ae9 itself does not
contain them.
Also recorded: 88 test node IDs disappeared and every one is accounted
for — 64 whole files whose subject was removed, 5 replaced by a better
file, 15 individually retired with their features, and 4 renames. No
meaningful coverage was lost, and the eight files that used a JavaScript
counter as instrumentation kept their assertions by moving the counter to
the world-state engine.
Six planning recommendations are reported, not applied. Three are marked
before M3: the threat model still describes the inherited SSRF guard's
opposite rule, TECHNICAL-DESIGN §5.1 still marks two hardening items
open, and the endpoint policy is a load-bearing security decision that
exists only as a module docstring and deserves an ADR.
M3 is clear to start. Its chokepoints are untouched or simplified — the
rollback paths now carry one shared state instead of two — and the Phase
0B undo/redo spike still applies. No M3 work here: Undo still deletes and
there is still no Redo.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017foPNqFjAJa2Ngebf5mEfL
94 files, +1,395 -6,578. Three files are new; twenty-four are gone. The
milestone is subtraction, and what is left is the single-user local
storyteller the specification describes.
Removed in full: campaign scripting and its QuickJS sandbox; multi-user
accounts, guest sessions, login, registration and the shared demo key;
the visitor-analytics tables, dashboard and page beacon; the access log
of sign-ins, addresses and devices; per-IP and per-user rate limiting
and quotas; Render deployment config; Postgres and psycopg; cloud
inference providers, the API-key field and the key encryption that
existed to store it; session-cookie signing. None of it was hidden
behind a flag — the routes are gone and answer 404.
Two things were kept that the brief allowed keeping. The `users` table
and its foreign keys stay as an internal ownership detail, because
rewriting them out means a migration across most of the schema to
delete a column that costs nothing; nothing creates a second user and
no request carries an identity. Five inert tables and four inert
columns stay for the same reason, so an M1 campaign database opens
unchanged.
The one addition is app/endpoints.py, which decides where a story may
be sent. Loopback, RFC1918, link-local, unique-local and CGNAT — an
explicit allowlist of networks, not a guess at what `ipaddress` means
by "private", which calls the documentation ranges private and IPv6
loopback reserved. Every address a hostname resolves to must be in it,
so a split answer does not squeak through, and the rule runs both when
the endpoint is saved and before every outbound request, because a name
that resolved to the LAN this morning can resolve elsewhere this
afternoon. Known cloud hosts are named in the refusal so the error says
why rather than looking like broken DNS. TLS is never traded against
it: M1's shared trust context is intact on all four clients and there
is no way to skip verification.
The hardcoded 120-second model timeout is now a setting. That was not
theoretical — on this GPU-less four-core host a cold load of
qwen2.5:3b-instruct took 648.9 seconds to produce the first turn, while
turns 2 to 5 of the same campaign took 3.6 to 13.1. Connect stays short
at 10s so a wrong address still fails fast; the read timeout defaults
to 300s and is bounded at 3600, because "wait longer" must stay a
number.
Two defects found while testing and fixed here. An unknown /api path
fell through the SPA catch-all and came back as HTML with status 200,
so a client asking for JSON parsed a web page instead of learning the
route was gone. And AIDND_CORS_ORIGINS accepted "*", which on an
unauthenticated loopback API would hand every page on the Internet a
write handle on the campaign database; it now refuses to start.
Verified rather than assumed. Offline, on a network with no route out
and no DNS: five turns, retry with both takes retained, restart with an
identical transcript digest, a failed model call leaving the accepted
AI-turn count untouched, and a capture with zero non-loopback unicast
packets. Against a real second machine on the LAN over HTTPS with a
private CA: four turns, restart, and a capture showing 289 packets to
the approved host, 344 loopback, zero anywhere else, zero DNS queries.
Cloud and public endpoints refused with their reasons; no API key
settable; every removed route 404.
604 backend tests pass, down from 648 by the fifteen retired with the
subsystems they tested and up by the twenty-nine added for the endpoint
policy and the removed surface. The scripting tests were not deleted:
eight files used a JavaScript counter as instrumentation for the state
snapshot and rollback machinery, which M2 does not touch, so the
counter moved to the world-state engine and those tests still assert
what they always did. Frontend lint and build are clean; the image
builds, and its wheel-building stage is gone with quickjs.
No M3 work. Undo is still destructive and there is still no Redo.
Phase 0B ran the upstream application on a network with no route out and
the first turn died in tiktoken, which downloads its BPE table the first
time anything counts a token. The browser separately fetched three font
families from Google on every page load. Neither is visible on a machine
that has been online once, which is why both now have tests.
The tokenizer table is vendored at
backend/app/context/vendor/cl100k_base.tiktoken and
backend/app/context/encoding.py builds the encoding from it directly,
verifying its SHA-256 against the digest tiktoken itself pins for that
URL. No code path in the tokenizer can reach the network any more —
not a warm cache, not an environment variable a deployment could forget.
The encoding was checked token for token against tiktoken's own.
The three font families are self-hosted as variable fonts under
frontend/public/fonts/ (343 KiB, Latin and Latin Extended), declared in
frontend/src/styles/fonts.css, and re-vendored by
frontend/tools/vendor_fonts.py. Their OFL licences ship beside them.
With no remote asset left, the CSP drops both Google hosts and gains
object-src, base-uri and form-action; woff2 also gets its real media
type, which Python's table lacks on a slim image.
A trusted-LAN Ollama turned out not to work at all over HTTPS. httpx
verifies against the certifi bundle, so an endpoint whose certificate
comes from a CA the user installed on their own machines — a StartOS
server's Ollama, for one — was refused with CERTIFICATE_VERIFY_FAILED
while curl and the browser on the same host accepted it.
app/tlstrust.py builds one context that unions the platform CA store
with certifi's, and all four outbound clients use it. A union rather
than a swap, so an image with an empty system store cannot start failing
on endpoints that worked before. Verification itself is untouched:
CERT_REQUIRED, hostname checking on, and no insecure escape hatch.
The storyteller listener is now loopback by explicit statement rather
than by inheriting uvicorn's default: start.sh, start.ps1, and
docker-compose.yml, which publishes to 127.0.0.1 rather than every
interface. Reaching an Ollama on another machine is outbound and needs
none of that inbound exposure.
backend/requirements.lock pins the exact tested closure;
requirements.txt keeps the ranges. DEVELOPMENT.md covers setup, the
same-host and trusted-LAN Ollama configurations, and how to re-run the
offline proof. PROVENANCE.md records the upstream commit, the MIT terms,
and both vendored assets.
Verified, not just compiled. On an --internal Docker network with
1.1.1.1 unreachable and no name resolving, a campaign was created and
played for six turns through same-host Ollama, restarted, and resumed.
A second run played ten turns through Ollama on a separate physical
machine on the LAN over verified HTTPS, summaries and embeddings
included, with the storyteller's default route deleted so the LAN was
reachable and the Internet was not. Its capture: 893 packets to the
approved host, 730 loopback, zero anywhere else, and zero DNS queries.
Two induced model failures left the accepted story bit-identical. The
inherited SPA was opened in a browser and a campaign read back from it.
Evidence is in planning/reports/M1-BASELINE-REPORT.md, along with the
findings that did not belong in this change.
648 backend tests pass, up from the inherited 632; frontend lint and
build are clean; the image builds. No M2 work is included: the hosted,
cloud, analytics, Postgres and scripting surfaces are untouched.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017foPNqFjAJa2Ngebf5mEfL
Establishes the production fork lineage decided in ADR 009. Upstream
AI-DnD is merged with --allow-unrelated-histories so this repository
carries the real upstream history alongside the planning package, and
future upstream commits can still be fetched and cherry-picked against
matching paths.
Upstream: https://github.com/parththakkar106/AI-DnD
Commit: d72f7c1bda
"Stop paying twice for a block a retry can still throw away"
Mon Aug 31 16:14:24 2026 +0000
tip of upstream/main at the time of the fork
License: MIT (c) 2026 Parth Thakkar — retained verbatim in LICENSE
No upstream file is modified by this commit; it is an import only.
Milestone M1 hardening lands on top of it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017foPNqFjAJa2Ngebf5mEfL
Validates the three finalists by clone, build, test run and live local
Ollama inference, then answers the fork question with measurements rather
than static review.
Recommendation: fork AI-DnD, confidence high. The Phase 0A call holds, but
it was wrong that AI-DnD's undo is non-destructive — retry preserves the
replaced take, undo hard-deletes it. A follow-up spike fixed that in 3
files (+130/-31): undo now moves a head cursor, redo round-trips, writing
below a moved-back head forks and keeps the abandoned line, branch-scoped
memory isolation survives, suite 627/632 with all 5 failures asserting the
deleted-row behaviour that was replaced.
Findings that change the plan:
- AI-DnD cannot take a turn air-gapped as shipped; tiktoken fetches its
encoding from a CDN. Proven on an internal Docker network, proven fixed
by vendoring the file.
- ai-adventure needs zero code for Ollama — two config lines — and its
turn/head/checkpoint schema is the target model to build to.
- Open Dungeon has zero automated tests and a positional summary
watermark, making its branch retrofit larger than Phase 0A costed.
- The world-state referee takes relative deltas; a 3B model sent absolute
values under full context, so a wounded player ended at full health.
Validation cannot catch this, so prefer ai-adventure's typed-event
vocabulary when generalising narrative state.
- Export/import recomputes head depth, so a round-trip silently undoes an
undo. Must be fixed alongside the undo work.
Docs only; no production code. Working tree from the runs stays untracked
under phase0b/.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015gUPLuxLs8wypxZPEmccJu
A memory whose block ends on the newest action is the one memory a player
can reach: retry and take-switching both refuse anything else. Each retry
of that turn withdrew the memory and wrote it again, and a block closes
every six actions while a normal turn writes two, so that was one turn in
three.
SETTLE_SLACK asks for one action past a block before the block is
summarized. The block is still MEMORY_INTERVAL actions; only the moment
moves. This is not the pre-SP4 holdback returning: that one was about a
retry rewriting text in place, which sibling attempts and forget_node
settled, and correctness still rests on the withdrawal rather than on the
slack. Undo and delete can carry a summarized node back to the tip, so the
withdrawal path stays reachable, just rarely.
The slack buys nothing back. The block that just closed is still in the
history window in full, so a memory of it says what the model can already
read.
Tests: the settling suite asserts the new rule and that a retry at the tip
finds nothing to withdraw; the rewrite suite builds thirteen actions so
both of its blocks settle; the one path test that ended on a block boundary
sets the slack to zero, because it is about which actions a block is read
from rather than about when a block forms. 632 green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015NcrxCJjqgDvAkamKeLWdn
The report opened with the connection string it was about to work on, taken
straight from AIDND_DATABASE_URL. On the hosted deploy that string carries the
Neon password, so the first line of every run put a live credential into the
console — and from there into scrollback, a screenshot, or a pasted bug report.
Nothing in the output said it was there to notice.
It now prints scheme, user, host and database name. The query string goes whole:
sslmode is the only part worth reading, and some drivers accept a password
there as well.
631 green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tqgupw5CZGjSZrUTNUd4fW
The report prints an arrow between the old and new word counts, and it prints
the memories themselves, which are model-written prose. A Windows console
defaults to cp1252 and cannot encode either, so the run dies on
UnicodeEncodeError partway through — after some memories have been rewritten
and committed, which is the worst place to stop. STATUS already carries the
same warning for tools/dbmeter.py.
stdout is reconfigured to UTF-8 with errors="replace" instead, so the report
degrades a character at a time rather than failing.
The usage notes were also written for a POSIX shell, and this project is
developed in PowerShell, where `VAR=x command` is not a thing. Both docs now
give the PowerShell form first, via Read-Host so the Neon URL and the secret
stay out of the history file, and name the venv's Python.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tqgupw5CZGjSZrUTNUd4fW
The hosted database is not the local one. It holds other people's stories, and
`summary_provider` builds from the adventure owner's Settings, so an unfiltered
`--write` against it would spend other people's money rewriting memories they
never asked about. `--adventure` could already hold a run down, but only if you
knew the ids.
`--email` names accounts instead, and every line of the report now says who owns
the adventure, so a dry run answers "whose keys would this spend" before
anything is written. Guests have no email and stay reachable only by id, which
is the right amount of friction for touching a stranger's bank.
The other half is written down rather than built: stored API keys are encrypted
with AIDND_SECRET_KEY, so a run from a checkout against the Neon database needs
the same value the web service holds. With a different one `decrypt_secret`
returns "" instead of failing, and every adventure is reported as having no key
— a run that looks like it worked and did nothing. The Dockerfile copies
backend/app alone, so tools/ is not on the box either way; the recipe is a
checkout pointed at AIDND_DATABASE_URL.
629 green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tqgupw5CZGjSZrUTNUd4fW
The prompt change only reaches memories written after it. plan/18 decided to
leave the existing ones alone and let eviction age them out at
memory_bank_capacity, on the grounds that re-summarizing would duplicate
whatever was still in the bank because nothing deletes the old rows.
That was wrong about the only option. A memory can be rewritten in place. The
row carries more than its text — whether it is pinned, how often it has been
retrieved, and the node it hangs off, which is what makes a fork inherit the
right memories — and rewriting `text` keeps all of it. Deleting the bank and
rewinding the cursor would lose that, and would trickle memories back at
MAX_MEMORIES_PER_RUN per turn, so an adventure nobody is playing would never
recover.
tools/rewrite_memories.py does it. Without --write it makes no model calls and
only reports the scope; --write rewrites, --embed re-embeds in the run rather
than leaving it to the app's post-turn pass. It reads whichever database the
app reads, so it works against the hosted Postgres as well as a local file.
Two things it needed from the app. `summarize_block` is now the one place a
memory prompt is assembled, and the post-turn pass calls it too — a backfill
that built its own prompt would be writing memories with a prompt that never
shipped, and nothing would report the drift. `source_block` reads a memory's
block back out of the story, which nothing has ever had to do: it reads on the
lineage of the branch the memory was written on, not the branch being played,
because after a fork the same depths hold different actions on each side and a
read through the adventure's path would summarize the wrong story silently.
It also excludes the discarded attempts at a retried turn, and tolerates a
block an action has since been deleted from.
Left alone: a memory with no source range, which is hand-written or migrated by
62 and may be the player's own words; a memory whose actions are gone; and an
adventure whose owner has no API key, because summarization spends the user's
own key by construction and never the shared demo key. --api-key/--model/
--endpoint override that, the last of them aiming a run at claude_shim.py.
The vector is cleared for every rewrite, because the stored one describes
wording that no longer exists. Re-embedding always uses the owner's own
embedding model, never --endpoint: a vector only means anything against the
vectors it is ranked beside.
17 tests, 627 green. The fork case is the one that would fail quietly, so the
test builds a fork whose depths hold different actions on each side.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tqgupw5CZGjSZrUTNUd4fW
The checked-in harness was run end to end through the shim on a newly
generated story, which is both a check that it works and an independent
replication.
The finding the change is for held: across two runs, none of the four
control memories names the protagonist and all four treatment memories
do. The length finding held too — controls at 34, 89, 105 and 107 words,
a four-fold spread with no budget stated anywhere, against 54 and 55
under MEMORY_MAX_WORDS.
One claim did not replicate, and the writeups now say so. Run 1 produced
two control memories in two different persons, which is the reported
complaint exactly, and plan/18 presented that as reproduced. Run 2's
controls were both "the player", consistently. Drifting between second
and third person is therefore something a model sometimes does, observed
once, not something it does every time. The naming gap is the durable
result and the writeups now lead with that instead.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NPyQN926gkZTAYfgugcaok
The run that justified MEMORY_MAX_WORDS lived in a scratch directory and
would have been gone with the container. The numbers in plan/18 were
therefore assertions nobody could check.
plan/18-appendix-memory-ab-run.md now carries the whole transcript: both
memories, both summaries, and the thirteen-action story they were
written from.
backend/tools/memory_ab.py reproduces it. It replaces the throwaway
script the first run used, and differs in two ways that matter. It goes
through OpenAICompatibleProvider rather than calling a model directly,
so a run exercises the provider, the streaming path and complete()
instead of a stub. And it reads the control prompt out of git at the
commit given to --before, so the thing being compared against cannot
drift from what actually shipped.
There was already a claude_shim.py serving an OpenAI-compatible endpoint
backed by the CLI, which is exactly what the throwaway script had
reinvented. memory_ab.py points at it by default, so a run spends a
Claude subscription rather than API credit, and --endpoint aims it at
the provider the deployed app really uses — which is the one question
this whole exercise could not answer.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NPyQN926gkZTAYfgugcaok
Ran the memory prompt end to end against a real model as a controlled
A/B: one story generated through the app's own build_context a turn at
a time, then both prompts run over the same blocks, so the story is held
constant and the prompt is the only variable. Fresh process per call, so
neither arm sees the other and the model is never told what is being
tested. The control is the exact prompt from 9cdcb55.
The reported fault reproduced. Two consecutive memories written from one
story minutes apart came back in two different persons — "You crept low
through the mist" and "The player asked Gwen to". With the cast brief
both named Kaelen. The control also inverted who acted on a move whose
player text was "grab her wrist and pull her down", which is the failure
the brief predicts: with no cast there is nothing to say whose wrist
"her wrist" was.
It also found something the prompt review had not. "1-2 plain sentences"
is not a length, and the same model wrote 34 words for one block and 105
for the next. A 105-word memory is a paragraph, and `memory_top_k`
injects five every turn, so the bank's running cost was set by a number
nobody had ever stated.
MEMORY_MAX_WORDS states it, and the prompt now says which details to
keep when trimming: the ones a later scene could turn on. Re-run over
the identical story, the same two blocks came back at 32 and 58 words,
still named, still third person, still carrying the camp map, the
strongbox behind the second tent, and the strap frayed near through.
Variance is the real gain — 34..105 became 32..58.
Overshooting 50 slightly is expected. Models exceed word budgets, which
is why builder.length_hint already carries a buffer for the same reason.
What this does not show: the run used a Claude model, and the app talks
to an OpenAI-compatible endpoint whose weaker models are why
worldstate/parse.py tolerates trailing commas. The prompt is followable
and the brief supplies the missing information; a weaker model is not
proven to comply as well.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NPyQN926gkZTAYfgugcaok
`_create_due_memories` sent six actions of second-person prose and
nothing else — no protagonist, no cast, no setting, and no instruction
about what person to write in. So for `You push the door open. She grabs
your arm.` the only honest memory was "You entered a room and she
stopped you", which names nobody when it is retrieved forty turns later.
The framing wandered too: with no rule, the model picked a person per
call, and one bank ended up holding "You entered the crypt", "The player
entered the crypt" and "He entered the crypt" for the same kind of
event.
Both prompts now carry a cast brief and a framing rule: third person,
the protagonist by name, other characters named rather than left as bare
pronouns. The rule states its reason, because a memory really is read in
isolation much later and a model told why complies far more
consistently than one handed a bare instruction.
The cast comes from the story cards, not from `stat_schema`. Every
schema NPC is already turned into a card at adventure creation,
deduplicated against the hand-written ones by name, so the cards cover
schema NPCs, an author's own cards, and an adventure with no RPG layer
at all through one path instead of three.
Keyword matching alone was not enough, and finding that out changed the
design. Built that way first, the brief for "She grabs your arm" listed
the protagonist and nobody else: the block that most needs a cast is
exactly the one written in bare pronouns, and Gwen's trigger keys
include "her" but the text says "she". So matched cards come first and
the remaining slots are filled with the other character cards. Places
and items are not topped up — an unmentioned tavern is not who "she"
was — though a place that is mentioned still matches normally. The
asymmetry with the turn prompt is deliberate: an untriggered card is
wrong as lore and right in a roster, because the roster answers "who
could these pronouns be" rather than "what is on stage".
Fixed descriptions only, never live values. `Gwen: trust 40 (wary)` in
the brief would make the same event summarized at two different times
come out framed differently, which is the fault this removes.
An adventure with no persona still gets the cast and the setting, and
the model is told to write "the player". One with nothing to say sends
byte for byte the prompt it sent before.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NPyQN926gkZTAYfgugcaok
Chromium against a fresh database with the demo scenarios seeded. No API
key needed: Insights assembles the prompt without calling a model, so the
checks run on the real assembled context rather than a stub.
21/21. The two request failures in the run were the sandbox — Google
Fonts is blocked by the egress policy, and the analytics beacon is
aborted on unload — not the app.
The case worth naming: a blank adventure with no RPG layer produced
sections `['narrator', 'persona', 'length_hint']`. That is what the
persona was added for, and it is the one thing no unit test in this repo
would have caught if the section had been gated behind `has_ws`.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NPyQN926gkZTAYfgugcaok
The player had stats but no identity. `stat_schema.player` carried hp and
mana beside `npc.gwen.trust`, but where an NPC has a name and a
description the player had neither, so the block rendered as
`You: hp 100/100` and nothing in the prompt said who "you" was.
Three columns on `adventures`: name, pronouns, description. All
user-only, all optional, and an empty name means the app behaves exactly
as it did before — no backfill, no special case for an adventure that
predates the migration.
They are adventure columns rather than part of `stat_schema` for two
reasons. An adventure with no RPG layer still has a protagonist, and
that is the case this was added for. And `worldstate.schema._initials`
treats every dict inside a stat section as a stat definition, so a
persona placed there would be instantiated, rendered in the guide, and
handed an `initial` value as though it were one.
The paths do not change. `player.hp` stays `player.hp`; only the label
moves, to `Kaelen (player): hp 100/100`, the same way NPC lines already
print a display name beside the id. A path carrying the persona's name
would break the moment a player renamed their character, because
`_history_text` replays every past turn's stored delta into the prompt
and those blobs hold literal `player.hp` strings.
The section sits in the system block. Only the user can edit it, so it
never changes mid-story and stays inside the cached prefix. That is what
makes it free, and it is why the AI must not be able to move it — a
delta aimed at `persona.*` is already refused by `_resolve`, and there
is now a test holding that in place.
The modal that used to appear only for scenarios with `${Placeholder}`
tokens now always opens, and is where the character is named. Persona
and placeholders stay independent: a scenario asking for `${Name}` is
asking its own question. No scenario in the repo uses placeholders at
all, so the overlap is hypothetical.
Phase 2, which feeds the persona and the cast to the summarizer, is
written up in plan/18 and not started. That is where the memory-quality
problem actually gets fixed; this change is what gives it a name to use.
Not yet driven in a browser — plan/18 lists what to check by hand.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NPyQN926gkZTAYfgugcaok
`place_action` used to read `depth` from `index`, so the fixture's retry
attempts all landed on the turn's own depth and formed one sibling group.
Now that depth comes from the head, each attempt moved the head one step
and every retry became a turn of its own. The fixture's own check caught
it: a coordinate holding a single superseded attempt has no live row, and
that turn disappears from the story.
The attempts at one turn share that turn's coordinate, so only the first
one goes through `place_action` and the rest copy its placement. This is
what `attempts.add_attempt` does. The fixture cannot call it directly,
because it makes the newest attempt live and the fixture needs a live
attempt that is often not the newest.
Also drop the `variant_index` and `variant_count` arguments, which raise
`TypeError` now, and read the two mark positions from a helper instead of
from the adventure columns migrations 72 and 73 removed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YFQY6WgaE3JaX3dXkxLynV
`seed_demo.py` still passed `index=0` to `models.Action`, which raises
`TypeError` now that the attribute is gone. Every call site inside `app/`
was updated when the column went, but the seed script sits outside the
package and was missed.
`bundle.settle` wrote `memory_cursor` and `summary_cursor`, which are no
longer mapped, so the assignments only set transient Python attributes.
The version 2 branch existed solely to make those assignments, and it ran
two queries per cursor to do it, so it goes. `settle` now returns early
when the bundle carries anchors. The `db` parameter is unused after that.
Also move the comment about deleted memories next to the
`db.delete(branch)` it describes. Splitting the router package left it
after the `finally`, where it read as attached to nothing.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YFQY6WgaE3JaX3dXkxLynV
`main` changed the four files this branch split into packages, so all four
came back as modify/delete conflicts. Each change is ported to the module
that now holds the code, unchanged in behavior:
- `delete_action` moves to `routers/adventures/actions.py`. It reaches the
turn lock as `turns.acquire_turn_lock` and `turns._active_turns`, which is
the rule this branch set: the package root no longer re-exports the names a
test rebinds.
- `EMIT_RULE` moves to `worldstate/parse.py`, and `_describe_stat` and
`render_reference` to `worldstate/render.py`.
- The chip-wrap rules move to `styles/schema-editor.css`, and
`StateChangeChips` to `pages/Play/reports.jsx`. `index.css` keeps this
branch's import list.
`test_delete_state.py` needed four edits to run here. It drops the eight-line
database prologue, because `conftest.py` does that once now. It imports the
shared `ScriptedProvider` from `fakes.py` instead of carrying a copy. It
patches `adventures.turns.OpenAICompatibleProvider` and reaches the lock the
same way. And it no longer passes `index=0` when it builds the start action,
because migration 71 drops that column.
564 tests pass. `npm run lint` and `npm run build` are clean, and the CSS
bundle is 56.71 kB, which is the pre-merge size plus main's new rules.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A chip's value carried `white-space: nowrap`, which was there so a signed
number never breaks between its sign and its digits. It was applied to every
value a chip can hold, and most of them are phrases rather than numbers: a
refusal reason, "no change — at its limit", and the whole new value of a
free-text stat, which is a sentence the AI wrote rather than a word.
An unbreakable phrase sets the chip's minimum width to the width of the
phrase, so `max-width: 100%` capped the box while the text kept painting past
the rounded border and off the screen. The chip row sits inside the story
text, which has nothing to scroll, so the page grew a horizontal scrollbar
instead: a 87-character status value measured 470px against a 320px phone.
The nowrap now applies only to the signed number, under `chg-num`. Every other
value wraps inside its chip, which is what the surrounding `overflow-wrap`
already did for the label.
Measured in Chromium at 320, 360, 390 and 430px wide: that same value paints
between 55px and 165px outside its chip before, and 0px after, with the page
no longer scrolling sideways. The number chip stays on one line.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017imUKVPhqophVUZZwJK7ST
The referee refused `npc.<id>.status` on a character that has a status stat
under another name, and the refusal was right: the model was reaching for a
path the guide never gave it.
The guide is the fixed list of everything a scenario tracks, and it named
stats in prose. A player stat and a world stat both read as a bare name, and
an NPC's stats read as the display name plus the stat — "Trainer Milo
active_status". The model had to turn that back into `npc.milo.active_status`
itself, and in the Pokemon demo five other characters carry a stat called
`status`, so the path it built was `npc.milo.status`. The live values do state
the paths, but only for the NPCs a scene has mentioned, so an NPC off screen
was addressable only by guesswork.
Each line now leads with the path: `player.potions`, `npc.milo.active_status`,
`flags.sandstorm_active`. An NPC's header is written even when it has no
description, because it is the one line that ties a display name to its id.
`EMIT_RULE` points at the guide for paths rather than at the live values
alone.
A free-text stat is marked `(free text)` beside its path, which is the wording
`EMIT_RULE` already used to describe it — the guide had been writing "free
text" at the end of the line instead. Such a stat now also gets a line when it
has no description, where before it was dropped and the model was left to send
a number for a stat that holds a string.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017imUKVPhqophVUZZwJK7ST
Deleting an AI reply removed the text and left everything the turn did to the
numbers standing. `script_state` and `world_state` belong to the adventure
rather than to the node, so the row going away takes nothing back. Undo, retry,
a take and a branch switch all restore them; this endpoint never did.
The visible symptom was the cooldown clock, which lives in
`world_state._meta.last_changed` and holds a depth. A deleted turn left its
depth there, and the turn played in its place is played at that same depth, so
the referee refused the change as one that had happened this very turn — on a
turn the story no longer contains. Delete the reply because you did not like
the stat change it proposed, press Continue, and the same change comes back
marked "changed too recently". A script's state stacked the same way: the gold
a deleted turn paid out stayed paid, and the replacement turn paid it again.
The restore reads the tip's own outcome rather than the deleted node's
neighbour, which is what `switch` does. Deleting the newest turn then rewinds
to the node in front of it, and deleting one from the middle of the story
restores the state the adventure is already in, so the text goes and the
numbers stay.
The endpoint now takes the turn lock too, for the reason undo takes it: it
writes the shared state, and a turn that is still generating is about to write
it as well.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017imUKVPhqophVUZZwJK7ST
Migrations 66 to 73 drop `actions.index`, `variants`, `variant_index`,
`variant_count`, `state_before`, and `world_state_before`, plus
`adventures.memory_cursor` and `summary_cursor`. `index` is a keyword in
SQLite, so migration 71 quotes it.
Nothing outside the migrations read these. `models.py`, `schemas.py`, and
`ACTION_LIST_COLUMNS` lose the same eight fields, `Adventure.actions` orders
by `id`, and `attempts.renumber`, `context.history.max_action_index`, and
`nodes.next_index` are deleted.
Two changes keep the migration replayable on a `create_all` database:
- `_split_variants_into_siblings` wrote through the live ORM table, so it
stopped compiling once migration 66 removed five of its columns. It now
writes through `_ACTIONS_AT_60`, a frozen `Table` with its own `MetaData`.
- Five data passes read columns these migrations drop. Each now calls
`_has_columns` and returns early when the columns are absent.
`bootstrap` takes a `through` version so a migration test can stop at the
schema it asserts on.
555 tests pass, up from 549. Eight of the new cases assert each column is
gone after a real schema-45 database migrates all the way.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0198qDK3gmgSo7EtQ4GTPqqK
Two items from Stage 2 of `plan/17-refactor.md`.
`frontend/src/hooks/useDebouncedSave.js` replaces three copies of the same
debounce. `PlotPanel` and `ScenarioEditor` held identical per-key timer maps.
`ScriptEditor` held a single shared timer, so editing two fields inside the
same 600 ms window canceled the first save. The hook gives every key its own
timer, which fixes that.
`Settings.stream` was dead state. Nothing read it and every turn streams. This
removes the column, both schema fields, and adds migration 65 to drop it. It is
item S1 in `docs/self-review.md`.
Migration 65 needs a new guard. `_column_already_gone` is the counterpart to
`_column_already_there`: `create_all` builds the current schema, which is
already missing every dropped column, so a fixture that stamps an old version
and replays would fail on a column that is not there.
Verified: 549 backend tests pass, lint and build are clean, and migration 65
runs both ways, once against a database that still has the column and once
against one that does not.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014Dix4oGV3njgWRdu7P9t6r
Stage 2, items 1, 4, and 5 of `plan/17-refactor.md`.
**One path resolver in `worldstate`.** `apply_delta` and `apply_override` routed
`flags.<name>`, `milestones.<id>`, `world.<stat>`, `player.<stat>`, and
`npc.<id>.<stat>` with parallel code, about 100 lines each. `_resolve` now says
what a path points at and returns either a target or the rejection to report.
Each function keeps its own write rule, because the rules genuinely differ: an
override sets a number rather than adding to it, ignores `cooldown`,
`max_delta_per_turn`, and the rule that a counter only counts up, and can un-set
a milestone.
A differential check ran both implementations over 3960 payloads: twenty paths,
fourteen values, three starting states, plus every three-path combination. The
results are identical except that 674 rejections from `apply_override` now carry
a `fix` string. `apply_delta` already worded those, and the world-state editor
renders them, so an override that names an unknown flag now explains itself the
way a delta does.
**`sse`, `SSE_HEADERS`, and `turn_error` move to `app/sse.py`.** Two routers
stream, and `chat.py` had to import from `routers.adventures` to reach them.
**`get_adventure_or_404` becomes the `current_adventure` dependency.** All 32
handlers repeated the call as their first statement. The ownership check now
reads in the signature and runs before the body. FastAPI caches a dependency for
one request, so the handler's `db` is the session the adventure came from.
The generated OpenAPI document is byte-identical except on `rename_branch`,
where `branch_id` is now listed before `adventure_id`, because that handler no
longer names `adventure_id` itself. Parameter order in the document is
cosmetic.
Six tests in `test_state_revert.py` call `undo_turn` and `retry_action`
directly rather than over HTTP. They pass the adventure they already hold
instead of an id.
549 tests pass.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014Dix4oGV3njgWRdu7P9t6r
`frontend/src/pages/Play.jsx` was 2280 lines holding 27 components. It is now
`frontend/src/pages/Play/`: `index.jsx` with the page component, five panels,
two drawers, the reports, the take pager, the refresh dialog, and the shared
formatting helpers.
Every line moved verbatim. A coverage check confirms every non-blank line of
the original appears exactly once, in order, across the twelve files, and a
name-resolution check confirms every identifier each file references is defined
or imported there, with no unused imports.
The line split stranded a comment at six of the boundaries. A leading comment
sits above the section it describes, so each cut left one at the end of the file
before it. All six moved to the section they describe, rewritten in the house
style.
`usePlaySession.js` is not here. The page component still owns all of the
session state. Moving eighteen `useState` calls and seven `useEffect` calls is a
rewrite rather than a move, and no frontend test would catch a mistake in it
today, so it waits for Stage 5.
Four stand-in providers under `backend/tools/` now define `last_usage = None`.
The turn engine reads that attribute after every call, and the fixtures never
defined it, so `tools.tree_fixture` crashed. That break predates this branch.
Verified by 549 passing tests, a clean `npm run lint` and `npm run build`, and
by driving the Play screen: the story view, all five panels, both drawers
including the world-state edit form, the branch map, the refresh dialog, and the
take pager stepping onto a take that lives on another branch. No console errors.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014Dix4oGV3njgWRdu7P9t6r
`frontend/src/index.css` held 2865 lines. It is now an import list, and each
section is its own file under `frontend/src/styles/`.
The order is unchanged. Two comments in the file already recorded that the
cascade depends on it: several selectors in the tome section win only because
they come after the base card rules, and the 720px query overrides the whole
desktop design. The built CSS bundle is byte-identical before and after, at
56686 bytes.
The plan asked for each `@media` block to move next to the rules it overrides.
That is not done, and should not be. Moving a rule past a later rule of equal
specificity changes which one wins, and there is no frontend test that would
catch it. The 720px block is now `responsive.css`, imported last.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014Dix4oGV3njgWRdu7P9t6r
`backend/app/worldstate/engine.py` held 918 lines covering four separate jobs:
reading a scenario's schema, parsing the block the model writes, applying a
change within the schema's limits, and rendering state as prompt text. Each is
now its own module, the largest 418 lines.
An AST comparison against the old file confirms all 29 definitions are
identical. No call site changes, because `worldstate/__init__.py` exports the
same names it did before.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014Dix4oGV3njgWRdu7P9t6r
`backend/app/routers/adventures.py` held 2353 lines and 35 endpoints. It is now
a package of 14 modules, the largest 443 lines.
The split moves text rather than rewriting it. An AST comparison against the
old file confirms all 86 definitions are identical, and the OpenAPI schema
still lists the same 35 operations.
Names a test replaces now live in `turns.py` only, and other modules reach them
as `turns.<name>`. Rebinding a re-exported alias changes the alias and leaves
every caller reading the original, so the package root does not re-export them.
A patch aimed at the old target raises `AttributeError` instead of passing while
doing nothing. Tests and the fixtures in `backend/tools/` say
`adventures.turns.<name>`.
The same rule keeps the turn lock working. One module owns `_active_turns`, so
one lock guards one set.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014Dix4oGV3njgWRdu7P9t6r
Every test module carried the same eight-line prologue redirecting the
database to a temp file. Only the first one to be imported ever took
effect: `app.database` reads `AIDND_DB_PATH` at import and builds `engine`
from it once, so by the time the second module ran the engine already
existed. The other 34 copies created a temp file that nothing opened and
nothing deleted, and leaked one per module per run.
`conftest.py` now does it once, which is early enough because pytest
imports conftest before any test module. It also deletes the file when the
run ends. The tests still share one database, exactly as they already did:
each `client` fixture calls `create_all` on setup and `drop_all` on
teardown, so no test sees another test's rows.
`tests/fakes.py` holds the one `ScriptedProvider`. Nine modules each had a
copy, and the copies had drifted into four feature sets, so a test that
needed to raise a provider error had to be written in one of the files
whose copy supported that. The shared one is the superset. The two
`FakeProvider` copies were the same class with a fixed reply, so they use
it too. `test_chat.py` keeps its own, which implements `chat` rather than
`generate` and records what it was constructed with.
An autouse fixture resets the fake's class state between tests, so a stale
reply list can no longer reach the next test.
435 lines out of the suite. 549 tests pass. Verified live by sabotage:
breaking the shared fake fails 13 tests across four modules.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014Dix4oGV3njgWRdu7P9t6r
Phase 17 splits the four files that hold most of the code, finishes the
schema migration SP8 left half done, and stops the published guide from
drifting away from its Markdown source. `plan/17-refactor.md` carries the
plan and the progress table, and `plan/STATUS.md` points at it.
Stage 0 is hygiene only. Both abandoned worktrees are gone, which freed
about 104 MB. Removing `sp7-tree-ui` needed one extra step: a Vite dev
server had been running out of it since 2026-08-18, holding
`frontend/.vite` open and owning port 5173, and serving a tree 54 commits
behind `main`. The three stale `.db` files are deleted; `data.db` is not.
`AIDND_TRUSTED_PROXY_HOPS` is now documented. It was read at `limits.py:55`
and named in no `.env.example`, README, or blueprint. It sets how many
proxy hops the rate limiter trusts in `X-Forwarded-For`, so a deployment
that adds a hop without setting it gets the bucket-rotation bypass back.
The 19 squash-landed branches are still there. `git branch -D` is blocked
by the permission classifier; the verified command is in the plan file.
549 tests pass.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014Dix4oGV3njgWRdu7P9t6r
`previous_titles` stops a rename stranding the row it left behind, but the rows
already stranded still had to be deleted by hand on every deployment. The
seeder now removes them on the next boot, which retires the stale "Road to the
Champion" demo without a database console.
Only rows with a NULL owner and `is_public` are considered, and a player's own
scenario is neither, so nothing anybody created is reachable. An adventure
started from a deleted demo survives: `adventures.scenario_id` is `ON DELETE
SET NULL`, and the adventure holds its own copies of the cards and scripts, so
it loses only the inherited cover art.
Two cases skip the sweep, because neither is an instruction to remove live
content: a seed file that fails to parse claims no title, and an empty seed
directory reads as a packaging failure.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PacdRuPXSkQQy4ZYdH32hF
An adventure has no cover art of its own and inherits its scenario's, and a
bundle carries no scenario id, because an id means nothing in another database.
The starter card therefore fell back to a monogram tile while the demo beside
it showed the Pokeball.
The starter file names its source under `scenarioTitle`, and the copy is linked
to the seeded scenario with that title. If no seed answers to the name, the
adventure keeps a NULL `scenario_id`, which is the state every imported bundle
is in and costs only the artwork.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PacdRuPXSkQQy4ZYdH32hF
`backend/tools/claude_shim.py` serves an OpenAI-compatible endpoint backed by
the local `claude` command line tool, so a demo can be played against a real
model with no API key. Each request spawns one `claude --print` process, which
suits the turn engine: the app assembles the whole prompt every turn and
expects a stateless endpoint.
Playing the Pokemon demo through it made all five of the world-state changes in
`plan/16` work, and the refusal loop ran end to end for the first time: turn 3
clamped to nothing, turn 4's assembled prompt carried the correction verbatim,
and the model's next delta was right. Two failures previously blamed on the
code were the demo model. One is a schema fault and is still open: Milo's three
Pokemon share one `active_hp` stat, so a switch leaves the newcomer at 0 HP.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PacdRuPXSkQQy4ZYdH32hF
An empty account gives a visitor nothing to read, and the daily demo turns are
limited, so learning what the app does cost one of them. `app/starter.py` now
copies a shipped export bundle into each new guest at the point the row is
created. The bundle is two exchanges of the Pokemon demo, which ends on a
knockout and shows an applied change, a refused one, and a milestone.
The guest row is committed before the copy is attempted, so a failure there
still leaves them with an account, and the copy runs inside a savepoint.
The row building that `POST /adventures/import` did inline moved into
`bundle.materialize`, which both callers use. The rate and size checks stayed
in the endpoint: the starter writes a file the server ships, so it has no
untrusted list to cap.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PacdRuPXSkQQy4ZYdH32hF
`seed.py` matches a seed file to its scenario by title, so renaming one
inserted a second public scenario and stranded the first. The stranded row
stays public forever and has to be deleted by hand on every deployment, which
is what happened to "Road to the Champion". A seed file now lists its old
titles under `previous_titles`, and the rename updates the existing row.
The cover art is a PNG data URI. `app/images.py` accepts raster formats only,
because SVG can carry script and the bytes are served from the app's own
origin, so an SVG stores but yields an empty `image_url`.
`tools/make_pokeball.py` draws the ball with `zlib` alone.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PacdRuPXSkQQy4ZYdH32hF
The model sent a total rather than a change for almost every number:
`npc.ivysaur.hp: 96`, `npc.milo.active_hp: 88`, `player.potions: 2`. Because
every `hp` starts at its maximum, each one clamped back to where it started, and
the potion count rose when the player spent one.
`EMIT_RULE` does say "deltas (not new totals)", but the scenario contradicted it
at closer range. `milo.active_hp`'s description said "Reset this to the
newcomer's full HP", which asks for an absolute and is injected every turn. The
bullets said "drop the HP, and raise it when healed", naming a direction but no
sign. Five HP descriptions said only whose HP it was. The one line that said
"not a delta" covered `player.active_pokemon`, so naming the exception made the
rule look optional. `pokemon_fainted` is the control: its description says "add
1 each time", and it is the only number that behaved.
Every stat description now states the sign, and the lead-in gives a worked
example. `milo.active_hp.max_delta_per_turn` goes 65 to 98, because a switch
moves that stat a full bar from 0 and the old cap made the reset unreachable in
one turn.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PacdRuPXSkQQy4ZYdH32hF
`world_delta_of` wrote `delta` and `applied` only. Every consumer that tells a
refused change from a successful one reads the two lists it dropped:
`Action.world_changes` marks a clamped chip from `clamped` and builds its
refusal chips from `rejected`, and `worldstate.refusals` reads both. So no chip
could report a limit, no rejection chip could appear, and no correction ever
reached the next prompt. The three mechanisms merged last session were live in
the code and unreachable in production.
Found by playing the Pokemon demo. Turn 3's snapshot held two clamped entries
with correct `fix` text, both chips came back `clamped: false`, and turn 4's
prompt carried no correction, so the model repeated the same mistake.
The 21 tests passed because `action()` built the column by hand with every list
present. It now fills the column through `world_delta_of`. Removing the two new
lines fails 9 of the 23 tests; that was checked by sabotage.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PacdRuPXSkQQy4ZYdH32hF
`plan/STATUS.md` pointed at SP9 as the next step and called it unmerged and
undeployed. SP9 and SP10 are both on `main`. They landed as squash merges, so
`git branch --no-merged` still lists their branches. The doc now points at SP8
and says to check `actions.parent_id` and commit `c0cd6fa` instead of the
branch list.
`plan/15-pokemon-demo-handover.md` named a cause for each of two bugs and both
causes were wrong. It carries a banner and inline corrections rather than a
rewrite, because the reasoning trap it fell into is worth keeping: the visible
evidence was "the number did not move", which reads as the model never trying.
`plan/16-world-state-refusals.md` records what the engine was actually doing,
the five changes, and a browser checklist. None of the UI work has been driven
in a browser.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PacdRuPXSkQQy4ZYdH32hF
`apply_delta` records three outcomes for every change the model sends:
`applied`, `clamped`, and `rejected`. Everything downstream read only
`applied`. A refused change reached the player as an ordinary chip, and
reached the model on the next turn as a change that had succeeded.
Five parts:
- `Action.world_changes` reads `clamped` and `rejected` beside `applied`.
Accepted stats carry a `clamped` flag; refusals become `kind: "rejected"`
entries. The `fix` key is present only when the engine wrote one, because
this property runs for every action of every list response.
- The UI separates the three outcomes. A clamp to a standstill reads
`no change - at its limit` on a dashed chip, a partial clamp is marked
`(limited)`, and a rejection carries its reason. Dashed and dimmed rather
than red: a refused change means the rules are working.
- The goals line names the milestone id, as `milestones.<id>`. The ids
appeared nowhere in the prompt before, so the model could not send one.
- Each rejection, and each clamp that moved nothing, builds a `fix` string
from the stat definition at the point of refusal. `render_refusals()`
renders them into the next prompt above `EMIT_REMINDER`.
- `_history_text` replays `applied_delta()` instead of the sent delta, so a
past turn's state block shows only what the engine accepted.
A clamp that reduced a change but still moved the value reports nothing. If
you tell a model its 80 damage became 30, it can treat the shortfall as a
debt and send the remaining 50 next turn, which is the swing
`max_delta_per_turn` prevents.
In the demo scenario, `pokemon_left` becomes `pokemon_fainted`
(`type: counter`, `initial: 0`). Starting at the ceiling turned a wrong-signed
delta into a silent no-op; counting up puts the wrong sign on the counter
rule, which refuses it out loud. The instructions also now ask for
`world.turn`, which sat at 0 for a whole playtest.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PacdRuPXSkQQy4ZYdH32hF
Each teammate is its own named entity with hp/status, the same shape
Milo already uses, rather than <name>_hp/<name>_status flattened
under player. Also replaces the five mutually-exclusive _active
flags with a single player.active_pokemon text field, mirroring
npc.milo.active_pokemon, so the AI no longer has to self-enforce
"exactly one flag true."
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PacdRuPXSkQQy4ZYdH32hF
