Round-4 review on PR #893 (no blockers, minor items):
- Document why Mistral's _is_transient is SDK-level only (429/5xx): a bare
connection drop the SDK surfaces as httpx/ConnectionError isn't an SDKError
and isn't retried here by design — the pod-rollover target is the gateway
(OpenAI-compatible) path, which does cover connection errors.
- Include the last error (%r) in the retry helper's "not resolved after N
attempts" error log.
- Add test_mistral_embed_batch_retries_on_5xx (batch path parity with embed()).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
From card 309 (OHR-Bench smoke-test triage): during a backend-pod rollover the
embedding endpoint was briefly unreachable, and openai.APIConnectionError /
ConnectError propagated unretried (the provider only retried 429). Documents
exhausted the 3 in-process retries and were dropped for that scan cycle.
Broaden the provider-level retry to the transient set -- APIConnectionError,
APITimeoutError, 429, and 5xx -- on the existing exponential backoff (2s->60s,
5 attempts), so a few seconds of retry rides through the rollover. Permanent
4xx (auth, bad request) still re-raise immediately. Generalize the shared
_retry helper (retry_on_rate_limit -> retry_on_transient, predicate renamed to
should_retry, accurate log label) with a back-compat alias; Mistral gets 429+5xx
for parity. The production gateway path inherits this via GatewayProvider, which
delegates to the decorated OpenAIProvider methods.
Add astrolabe_vector_ingest_dropped_total{reason}, incremented when a document
exhausts retries, classified (connection|timeout|rate_limit|server|qdrant|other)
by _drop_reason so the embed-drop rate is alertable per cause. Dropped docs are
NOT marked failed, so the next full scan re-picks them (re-queue via scan loop).
Refs: Deck board 12 card 309 (AC #1 no permanently-dropped docs; embed-drop
metric for AC #5).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Billing product model finalized (Deck #281): bill pages externally, record
tokens internally. Rename the data-plane metric literals to match the now-
canonical contract (Deck #284) — the control plane's METRIC_EVENT_NAMES is
already renamed, so the old names would be unmapped and never sync to Stripe.
Rename (values unchanged):
- embeddings_queries → tokens_embedded (value = real token count, already
emitted by this PR; the unit upstream providers bill on).
- pages_chunks → pages_embedded (value kept as len(chunk_texts) interim;
TODO(#282): real normalized "pages indexed" count — real pages for paginated
types, chars/tokens-per-page constant otherwise — is deferred to the
instrumentation card, this only lands the name/contract).
- All literals, log strings, docstrings, comments, the migration comment, and
tests renamed; grep confirms zero old strings remain.
Observability (new): export embedding token cost to Prometheus as
astrolabe_embedding_tokens_total{provider,operation} (operation = index|query)
so the billed cost unit is visible in Grafana, not just the per-tenant billing
DB. Dedicated counter (doesn't inflate the existing chunk/request metrics) and
always-on (independent of USAGE_METERING_ENABLED, so OSS/self-host gets it).
Wired on both the indexing batch embed and the search query embed (query inside
the per-request cache-miss branch, so reused embeddings aren't double-counted).
Note: the rename orphans any pre-existing embeddings_queries/pages_chunks rows
in tenant app DBs (CP no longer maps them) — acceptable; pipeline is inert with
throwaway dev/sandbox data.
Deck #284 (folded into PR #875).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Round-1 claude-review findings:
- 🔴 Multi-doc_type search billed N embedding calls as 1. nc_semantic_search
loops search() once per doc_type on one BM25HybridSearchAlgorithm instance,
and each call re-embedded the query, so only the last query_token_count was
recorded. Cache the dense embedding per query on the (per-request) instance
so the query is embedded — and metered — exactly once regardless of how many
doc_types are searched. This also removes the redundant per-type embed work
and avoids billing a user N× for one logical query.
- 🟡 Ollama embed() now delegates to embed_with_usage() so single and batch
embeds use the same /api/embed endpoint (was the legacy /api/embeddings),
keeping _detect_dimension and other embed() callers consistent.
- 🟢 round() instead of truncating int() when coercing provider-reported token
counts (forward-compatible if a provider ever returns a float).
Tests: per-instance query-embedding cache (embedded once across 3 doc_types;
re-embeds on a different query).
Deferred (stated on the PR): mistral/openai single-embed dual path (changes
tested error/request semantics on the cloud-critical path — separate refactor),
bedrock boto3 sync-in-async (pre-existing; no new invoke_model calls per doc).
Deck #67.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
embeddings_queries now records the embedding request's token count (the unit
upstream providers bill on) instead of an operation count, and fires on the
indexing path too. Previously only semantic search recorded it (value=1), so a
re-indexing run produced no embeddings_queries events at all — only pages_chunks.
- Provider layer: additive embed_with_usage / embed_batch_with_usage surface the
per-request token count (Mistral/OpenAI usage.total_tokens, Bedrock Titan
inputTextTokenCount, Ollama prompt_eval_count); a char-based estimate is the
fallback (Simple, and any provider/response without a token field). Gateway and
EmbeddingService forward through. The count travels as a return value / a
per-request SearchAlgorithm attribute — never on the singleton — so concurrent
indexing + search can't mis-attribute bills.
- Indexing (vector/processor.py): records embeddings_queries (value=batch tokens)
alongside the existing pages_chunks event.
- Search (server/semantic.py): value is now the query embedding's token count,
relayed from BM25HybridSearchAlgorithm via query_token_count.
The astrolabe_embeddings_queries Stripe meter (sum aggregation) now sums tokens
with no CP/Terraform change. The meter "queries"->tokens naming/unit
clarification (homelab-terraform #254) + CP rollup/portal copy is a follow-up.
Deck #67.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- _retry.py: replace `assert last_error is not None` with explicit
`if last_error is None: raise RuntimeError(...)` so the original
rate-limit error is preserved under `python -O`.
- openai.py: drop the `_retry_factory` alias chain; rename the bound
decorator to `_retry_429` to match the pattern in mistral.py.
- mistral.py: comment the imports so future reviewers understand why
`from mistralai.client import …` is the canonical path on 2.x (no
top-level `__init__.py`; no `mistralai.models` subpackage either).
- docs/configuration.md: add `OPENAI_GENERATION_MODEL` and
`OLLAMA_GENERATION_MODEL` rows to the env-var reference table.
- test_mistral.py: add direct unit test for the `_is_rate_limit`
predicate (429 → True, 500 → False, missing-attr → False).
- test_registry.py: stub `mistralai.client.Mistral` in the registry
picker test, mirroring the Ollama sibling, so the test doesn't
depend on the SDK accepting arbitrary keys.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Addresses the Claude Code review on PR #772 plus the SonarCloud S1192 finding:
- Extract `retry_on_rate_limit` into `nextcloud_mcp_server/providers/_retry.py`
as a parametric decorator. OpenAI and Mistral now share the same backoff
loop; future providers can reuse it without copy-paste.
- New `tests/unit/providers/test_retry.py` covers the decorator: 429 retry +
success, non-429 immediate re-raise, MAX_RETRIES exhaustion, default
predicate, and unrelated exception passthrough.
- Tighten Mistral SDK import to `from mistralai.client.errors import SDKError`
(the canonical sub-path; the reviewer's `from mistralai.models import
SDKError` does not exist in mistralai 2.4.5).
- Replace `MistralProvider.close()`'s direct `__aexit__` call with a no-op +
comment — the Speakeasy-generated client has no public close hook and the
underlying httpx client is closed by GC.
- Extract the duplicated "Embedding not supported" message to a module-level
constant (SonarCloud S1192).
- Align `Settings.get_embedding_model_name()` Bedrock check with the registry
by also considering `bedrock_generation_model`.
- Add the `mock_mistral_client` fixture to
`test_mistral_no_embeddings_disabled` for parity with the rest of the file.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a hosted Mistral embedding option (mistral-embed, 1024-dim) alongside
the existing Bedrock / OpenAI / Ollama / Simple providers. Implementation
mirrors OpenAIProvider: lazy dimension detection with a known-models lookup,
chunked batch requests, defensive index sort, and a 429-aware retry decorator.
In the same change, ProviderRegistry switches from os.getenv to the
dynaconf-backed Settings dataclass so all five providers share a single
configuration path. config.py gains the previously-uncovered Bedrock keys,
the new Mistral keys, the missing OPENAI_GENERATION_MODEL /
OLLAMA_GENERATION_MODEL, and SIMPLE_EMBEDDING_DIMENSION.
Auto-detection priority: Bedrock → OpenAI → Mistral → Ollama → Simple.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
