feat(summarizer): resilient endpoint chain with local→cloud fallback (ADR-022)

The first friendly-pilot live run produced zero summaries: koala/phi4-mini hit
three silent failure modes — 8k context overflow on long transcripts (HTTP 400),
intermittent malformed JSON (highlights as a bare string), and no fallback wired
at all (summarizer.New(primary, nil)).

Keep phi4-mini as the fast primary and add resilience around it:

- Ordered endpoint chain (summarizer.NewChain): phi4-mini → koala/phi4-14b
  (local) → berget/mistral-small (worst-case external). All reached through the
  one LiteLLM gateway by alias.
- A parse failure now advances the chain like a transport error — the old
  Primary→Fallback shape returned the parse error without trying anyone else.
- Tolerant parse: highlights/takeaways coerce string→[]string, absorbing the
  common small-model quirk without spending a fallback round-trip.
- Transcript truncation (TAPIR_MAX_TRANSCRIPT_CHARS=18000) prevents the overflow
  rather than recovering from it; validated to fit phi4-mini's 8k window.
- Bounded completion budget (TAPIR_SUMMARY_MAX_TOKENS=1500) — the old 8192 budget
  itself contributed to the overflow.

Local-first guarantee preserved by ordering: external endpoint is tried only
after every local one fails. TAPIR_CLOUD_FALLBACK_MODEL="" disables it entirely
for client/NDA deployments.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

DECISIONS.md

@@ -803,6 +803,58 @@ never exposed in the UI.
 
 ---
 
+## ADR-022 — Summarizer is a resilient endpoint chain, not a single model
+
+**Status:** Accepted (2026-06-10). **Extends ADR-004** (the copied `llm` Primary→Fallback
+routing). Triggered by the first friendly-pilot live run, where a connected user got **zero**
+summaries after 12h.
+
+**Context.** Stage-0 ran a single summarizer model (`koala/phi4-mini`) with no fallback wired
+(`summarizer.New(primary, nil)`). The live run exposed three independent failure modes, each of
+which silently produced no summary:
+
+1. **Context overflow.** `phi4-mini` has an 8k context. Real transcripts (one was 11,602 tokens)
+   exceed it and the gateway returns HTTP 400 — and the request also sent `max_tokens=8192`, so
+   even a short transcript plus the completion budget could overflow the window.
+2. **Malformed model output.** `phi4-mini` intermittently emits `highlights` as a bare string
+   instead of an array, producing `cannot unmarshal string into []string`. The old code returned
+   the parse error **without** trying any other model — a 200-with-bad-JSON short-circuited.
+3. **No fallback existed at all** — any primary failure was terminal for that video.
+
+`phi4-mini` is kept as primary deliberately: it is fast and, on transcripts that fit, correct.
+The fix is resilience around it, not replacing it.
+
+**Decision.**
+1. **Ordered endpoint chain (`summarizer.NewChain`).** Endpoints are tried in order; the first to
+   return a *parseable* summary wins. Default chain:
+   `koala/phi4-mini` (primary, local) → `koala/phi4-14b` (fallback, local) →
+   `berget/mistral-small` (worst-case, external). All three are reached through the **one** LiteLLM
+   gateway by alias — the gateway already fronts both llama-swap and berget — so a fallback is a
+   different alias, not a second client config.
+2. **A parse failure advances the chain, same as a transport error.** "Reliably summarized" means
+   *parseable summary returned*, not *HTTP 200*. This is the behaviour the old Primary→Fallback
+   shape missed.
+3. **Tolerant parse.** `highlights`/`takeaways` coerce from a bare string (or a mixed scalar
+   array) to `[]string`, so the most common small-model quirk is absorbed **without** spending a
+   fallback round-trip — keeping the fast path fast.
+4. **Transcript truncation (`TAPIR_MAX_TRANSCRIPT_CHARS`, default 18000).** Input is bounded
+   up-front to fit a small-context primary, so overflow is prevented rather than recovered-from.
+5. **Bounded completion budget (`TAPIR_SUMMARY_MAX_TOKENS`, default 1500).** A summary needs few
+   hundred tokens; the old 8192 budget itself contributed to 8k-window overflow.
+
+**Local-first guarantee preserved.** The chain ordering *is* the guarantee: locals are tried
+first, so content reaches the external endpoint only after every local endpoint has failed.
+`TAPIR_CLOUD_FALLBACK_MODEL=""` removes the external endpoint entirely — the lever a
+**client/NDA deployment** pulls so content never leaves the local stack. With no external endpoint
+configured the `ai_routing.feature` "content only local" scenarios hold unchanged.
+
+**Reversibility.** Pure wiring + config. Setting `TAPIR_FALLBACK_MODEL` and
+`TAPIR_CLOUD_FALLBACK_MODEL` empty collapses the chain back to single-primary behaviour; the
+tolerant parse and truncation are strict supersets of the old behaviour (a previously-parseable
+reply still parses; a transcript within budget is unchanged).
+
+---
+
 ## Rejected alternatives
 
 Approaches considered during the 2026-06-02 planning + grill session and **deliberately not