Add three per-user AI settings, all configured in the app (no new env vars),
behind one migration (0198):
- Response timeout (seconds). A per-user upstream timeout threaded onto the
provider call. A local/self-hosted backend can take more than the legacy
60 s default to load a model on the first request, which timed the Coach
out before the first token. Defaults to a generous ~180 s when unset;
validated to 10–600 s.
- True token streaming for the local OpenAI-compatible client. The client now
POSTs with stream:true and parses the SSE delta chunks, emitting tokens as
they arrive, with a per-idle-gap timeout instead of a whole-call ceiling and
a graceful fallback to the buffered path when a server cannot stream. The
Coach route runs the provider call inside the response stream behind a
keepalive heartbeat so a reverse proxy no longer idle-drops the connection
during model load or generation. Grounding, refusal, and sentinel-stripping
still run on the complete reply before any token reaches the client.
- Dedicated document-scan (Lab-OCR) provider. Off by default — the OCR path
uses the main provider chain. When enabled, lab-report processing resolves a
separate provider/model/key, with a hint when the chosen model is not
vision-capable. The new key column is AES-256-GCM at rest and registered in
the encrypted-columns registry and the rotation script.
i18n: EN + DE translations for the new Settings → AI strings (other locales
carry interim copies to keep locale parity green).
A Codex backend 400 on the richer Coach tool-call request (and the daily
briefing) propagated out of the provider chain runner unwrapped: a 4xx other
than 401/403/429 was not classified as a hard provider failure, so it bypassed
AllProvidersFailedError and rethrew from the chat route as an HTTP 500. Plain
status-card completions, which never send a tools array, were unaffected.
Harden the failure path so a provider error can never surface as a 500:
- Treat any 4xx a provider wire returns as a hard failure so it cascades and is
wrapped in AllProvidersFailedError. InsightSchemaError (the strict-JSON 422
surface) stays the only non-hard class, so that diagnostic is preserved.
- The chat route now also classifies a tagged provider error that reaches its
catch unwrapped and emits the graceful coach.provider.* SSE frame instead of
rethrowing. Only an untagged error keeps the 500 + error-reporter path.
Make the Codex 400 diagnosable: fold the redacted upstream response body into
the thrown error message and the chain wide-event, so the field the backend
rejects is visible in logs (it was previously dropped).
Fix the Codex request shape behind the tool-call 400:
- Send the required strict flag on each Responses-API function tool.
- Replay prior assistant turns as output_text, not input_text, per the
Responses API multi-turn contract — the violation only surfaced on round two
of the tool loop, which is the path that broke.
Codex-only; the OpenAI, Anthropic, local and mock providers are untouched.
Each retrieval tool re-scoped the snapshot to a single source, so its cache key
never matched the inventory's full-source build — a four-tool turn paid four
extra full snapshot builds on top of the inventory's. Thread the inventory's
probe scope to every tool so a per-tool read lands the one cache entry the
inventory already primed, collapsing the turn to a single build (a window
override still gets its own correct build). Raise the loop's round cap from two
to three so a sequential cross-metric why-chain isn't starved; the per-round
token budget is unchanged and the absolute ceiling tracks one above so the
final round still forces prose.
The daily AI-token gate was a flat 25,000/day for every provider. That cap
exists to bound the OPERATOR's API bill, but a ChatGPT-OAuth (Codex) or BYOK
chain egresses on the USER's own plan/key and costs the operator nothing, so
gating it on the operator-cost ceiling is wrong. A single gpt-5.x reasoning
turn legitimately reports 20k-40k gross total_tokens (re-sent system prompt +
inventory + tool defs per round + hidden reasoning, summed across rounds), so
one or two turns exhausted the flat cap and locked the user out of a plan they
pay for.
Classify the chain's cost owner by its primary provider via resolveDailyCap:
an admin-openai primary (operator pays) keeps the operator-cost cap; every
user-egress primary (codex/openai/anthropic/local) gets a generous
abuse-only ceiling. Threaded into the Coach chat gate and both OCR-extract
modes.
Raise the operator-cost cap 25,000 -> 200,000 so the operator-key path also
survives a normal day of reasoning turns, keeping gross-token accounting
rather than reworking the reconcile math.
Subtract cached input tokens at reconcile: the Responses-API gross total still
includes prompt-cached input the user did not re-pay for. The codex client
already parsed cached_tokens; sum it through the tool loop and bill
total_tokens minus cached.
The Daily Briefing strips any number absent from its server-computed
figures; the Coach's tool path had no equivalent, so a transcription or
paraphrase drift (a tool returns systolic 128, the reply says ~138) could
ship. Collect the numeric leaves from every present tool result this turn,
extract the numbers the reply asserts, and soft-correct any that match no
fetched figure to a neutral placeholder — annotated, non-blocking, and a
no-op on a qualitative turn or the no-tools path where there is nothing to
grade against. The bounded retrieval loop now returns the present results'
payloads so the check has an authoritative figure set.
Replace snapshot-stuffing in the Coach turn with a small base context plus
on-demand retrieval. The model receives the full system prompt, a tool-mode
grounding addendum, and a tiny DATA INVENTORY manifest listing which domains the
user has logged and which tool fetches each; the figures themselves are pulled
only when a question needs them.
Six read-only tools wrap the existing server-authoritative snapshot builder:
get_metric_series (blood pressure / weight / pulse + the additive series),
get_glucose_panel, get_sleep, get_medication_compliance, get_labs, and
get_illness_recovery. Each returns a structured { present: false } for an absent
or opted-out domain rather than throwing or returning an ambiguous empty list,
so the model can always tell missing data from real data. Module and cycle gates
are enforced at the builder, so opted-out domains are unfetchable.
A bounded loop runs the call -> tool-result -> answer cycle: at most two rounds
with a hard cap of three that forces a final answer, parallel tool execution,
and only the final round streamed. The day's budget reserves the per-call
ceiling times the round count and reconciles the summed tokens; the atomic
reserve and reconcile primitives are unchanged. All inbound gates (rate limit,
refusal, replay scan, consent, budget) stay before the loop.
The legacy snapshot path stays alive as the no-tools fallback: tool mode is on
only when every provider in the chain supports tool-calling, so a chain that
includes a local model falls back to the prior behaviour verbatim. The tool
trace is persisted onto the assistant message provenance.
Replace the CompletionParams { systemPrompt, userPrompt } string pair with a
structured { system, messages[] } request across all five AI clients. One
refactor unlocks three things: provider prompt-caching (Anthropic cache_control
on the stable system prefix, OpenAI/Codex automatic prefix cache, with
cachedInputTokens surfaced for observability), the tool-calling wire surface,
and clean multi-turn.
Each client maps the new shape to its own wire and degrades unsupported
features silently: the local OpenAI-compatible server never receives a tools
field (most reject unknown keys), and prompt-caching is best-effort everywhere.
CompletionResult gains toolCalls, finishReason and cachedInputTokens; the tool
types and per-provider mapping ship now, the tool-use loop does not.
Vision inputs fold into the user turn's content parts via singleUserTurn, with
each provider's pre-existing block ordering preserved. The corrective-retry
paths append to the last user message rather than a flat string. The Coach
keeps assembling one user turn so its snapshot-once and grounding behaviour is
unchanged. No wire schema visible to API consumers changed.
Behaviour is unchanged across every surface; the full suite stays green.
The full SNAPSHOT block (every cluster plus labs, illness, cycle and the
reference-grounding lines) was interpolated into the prompt on every turn,
so a one-word follow-up re-paid the entire grounding prefix. Ship it only on
the first turn, or once the history window starts eliding turns; on a cheap
follow-up send a short pointer back to the snapshot the model already received
earlier in the same conversation. The prior turns ride the transcript, so
grounding and cross-metric correlation are preserved while a follow-up turn
sheds about 6,000 tokens. Record snapshotSent and promptChars on the reply
annotation so the cut is observable.
One-shot prettier pass so the new format:check gate starts from a clean
baseline. Formatting only — no behaviour change. Generator-owned and
vendored artifacts stay untouched via .prettierignore.
Per-route tests assert the two-layer Coach module gate: a disabled
module (operator availability OR per-user disableCoach) yields 403
module.disabled before any Coach work runs, and an enabled module falls
through to the normal path. Covers the chat POST + list GET, about-me
GET/PUT, the questions GET/DELETE, and the adopt POST short-circuit
ahead of the locked read-modify-write.