A session created under the openai-compatible provider and resumed under the
anthropic-compatible provider (same vLLM model) failed with an opaque
InternalError instead of recovering. Root cause: vLLM returns a context-window
overflow as HTTP 400 BadRequestError on /v1/chat/completions but HTTP 500
InternalServerError on /v1/messages, and the rehydrated resume payload overflowed
the window. The 500 was retried four times then surfaced as a bare class name.
- Detect overflow by message text, not exception class (_is_ctx_overflow),
shared across the fatal-error formatter, both stream-retry gates, the send-loop
recovery, the chunker, and the task_agent loop. Overflow is non-retryable
(deterministic; no backoff). Phrasing is overflow-specific so a token-quota
rate-limit isn't misclassified.
- Proactive pre-send compaction (Layer A): when already over the hard ceiling,
compact once before the first stream so a resume that arrives over-window (or
follows a switch to a smaller-context model, with no prior compaction) doesn't
go out blind. Generation-guarded end to end so an orphaned or superseded send
can never swap the live generation's history.
- Binary-subdivision chunker: an over-window summary batch is split in half and
the partials merged (~log2(N) calls, not one per block); a lone over-window
block is truncated progressively down to a floor before bailing irreducible.
- Cooperative cancellation honored through compaction; send() consumes its own
generation's cancel signal on exit, so a stale cancel can't block a later
idle /compact and a live cancel is never disarmed.
- _format_backend_error surfaces "Context window exceeded ..." instead of an
opaque InternalServerError, and only for unrecognized classes.
- retry/rewind, the continuation hint, and title generation all exclude the
synthetic [Conversation summary] turn so they can't target the label.
- task_agent salvages a sub-agent's partial work on any terminal error (not only
overflow), re-raising only when there is nothing to salvage.
A bare ``httpx.ReadTimeout`` previously surfaced as ``ReadTimeout: timed
out`` — no provider, no base URL, no model — leaving the user with no
signal to tell whether a model server hung, the URL was wrong, or the
model isn't loaded on the backend.
``ChatSession._format_backend_error`` now rewrites known boundary
exceptions (httpx ``ReadTimeout`` / ``ConnectError`` / etc. and OpenAI /
Anthropic SDK ``APITimeoutError`` / ``APIConnectionError`` /
``NotFoundError`` / ``AuthenticationError`` / ``RateLimitError``) into
operator-actionable text that names the provider, base URL (query
string stripped before ``sanitize_error_text`` redacts credentials),
and model. Matching is by class name so the helper carries no SDK
imports. Unrecognised exceptions fall through to the legacy
``f"{type(exc).__name__}: {exc}"`` shape, preserving existing grep
targets.
Patrick Buckley