* feat(claude-code): full-session retain mode with document upsert and configurable tags
Switch default retain behavior from per-turn chunks to full-session upsert.
Each session is now retained as a single document (document_id = session_id)
that gets updated on every Stop event, instead of creating fragmented
documents with timestamp-suffixed IDs.
New config options:
- retainMode: "full-session" (default) or "chunked" (legacy)
- retainTags: list with template variable support ({session_id}, {bank_id}, {timestamp})
- retainMetadata: extra metadata dict merged with built-in fields, supports templates
* fix(claude-code): respect retainEveryNTurns in full-session mode
The turn-count gating was only applied in chunked mode, meaning
full-session mode would re-ingest the entire transcript on every
single Stop event. Now retainEveryNTurns gates both modes.
Also fix test isolation: resolve ~/.hindsight/claude-code.json at
call time (not module load) so HOME override in tests works correctly.
* fix(claude-code): fix config tests after USER_CONFIG_PATH removal
Update tests to use HOME env var override instead of monkeypatching
the removed USER_CONFIG_PATH constant. Add autouse fixture to
TestLoadConfig to isolate all config tests from real user config
and HINDSIGHT_* env vars.
* docs: add supported platforms section and Windows installation guide
Adds a platform compatibility table (Linux, macOS, Windows) and a
dedicated Windows setup section with step-by-step instructions for
installing PostgreSQL + pgvector and running Hindsight natively.
Follows up on #699 which added Windows native support.
Also fixes a ty type-check error in metrics.py for the conditional
resource module import.
* chore: sync generated clients and lock file after #699
Regenerate client SDKs to pick up ValidationError model changes
and update uv.lock with platform-specific uvloop/winloop deps.
* docs: update Windows section — pg0 now supports Windows
pg0 v0.12.0 added Windows support, so embedded DB works everywhere.
Restructure Windows section to show simple install-and-run first,
with external PostgreSQL as an optional alternative.
* chore: sync generated docs skill and openapi references
FastAPI generates the ValidationError schema with only loc, msg, and
type, but Pydantic v2 actually returns input, ctx, and url as well.
Generated clients with strict JSON decoding (Go's DisallowUnknownFields)
cannot parse real 422 responses — the actual validation message gets
replaced by a confusing JSON decoding error.
- Patch the OpenAPI schema in create_app() to add input, ctx, url
- Regenerate spec and Go client
* feat: Windows native support — run Hindsight without Docker on Windows
Four compatibility fixes that allow Hindsight to run natively on Windows
with an external PostgreSQL + pgvector installation:
1. **pyproject.toml**: Conditional event loop dependency
- `winloop` on Windows (sys_platform == 'win32')
- `uvloop` on Linux/macOS (sys_platform != 'win32')
2. **main.py**: winloop integration via `winloop.install()`
- Patches asyncio event loop policy globally before uvicorn starts
- uvicorn sees "asyncio" but runs winloop underneath (same perf as uvloop)
- Falls back to default asyncio if winloop unavailable
3. **metrics.py**: Guard `resource` module import
- `resource` is Unix-only (getrusage, getrlimit)
- Conditional import with None fallback
- Skip process metrics collection on Windows
4. **fact_storage.py**: Cross-platform strftime
- `%-d` (no-padding day) is glibc-only, fails on Windows
- Replaced with `%d` + `.replace(" 0", " ")` for same output
## Windows Setup Guide
### Prerequisites
- Python 3.11+
- PostgreSQL 17 with pgvector extension
- Ollama (for local embeddings) or external embedding provider
### Install PostgreSQL + pgvector on Windows
```bash
winget install PostgreSQL.PostgreSQL.17
# Build pgvector from source (requires Visual Studio Build Tools)
git clone https://github.com/pgvector/pgvector.git
# In x64 Native Tools Command Prompt:
set PGROOT=C:\Program Files\PostgreSQL\17
nmake /F Makefile.win
nmake /F Makefile.win install
# Enable extension
psql -U postgres -d hindsight -c "CREATE EXTENSION IF NOT EXISTS vector;"
```
### Install and Run Hindsight
```bash
pip install -e ".[embedded-db]"
# Set environment variables
set HINDSIGHT_API_LLM_PROVIDER=openai
set HINDSIGHT_API_LLM_API_KEY=your-api-key
set HINDSIGHT_API_LLM_BASE_URL=https://your-llm-endpoint/v1
set HINDSIGHT_API_LLM_MODEL=your-model
set HINDSIGHT_API_DATABASE_URL=postgresql://postgres@localhost:5432/hindsight
set HINDSIGHT_API_EMBEDDING_PROVIDER=ollama
set HINDSIGHT_API_PORT=8889
hindsight-api
```
Data persists in PostgreSQL on your local disk — survives reboots,
updates, and anything that would wipe a Docker volume.
Tested on Windows 11 with PostgreSQL 17.9, pgvector 0.8.2,
Python 3.11, RTX 5080 (CUDA embeddings + reranking).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: handle strftime ValueError on Windows in fact_storage
The strftime call on occurred_start/occurred_end can raise ValueError
on Windows when the datetime object has unexpected format properties.
Wrap in try/except to gracefully skip date signal rather than crash
the entire retain batch.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: control plane UI fixes for recall and data view
- Sanitize NaN cross-encoder scores to 0.0 in reranking pipeline
(Pydantic serializes NaN as JSON null, breaking UI score display)
- Add null-coalesce for score in search debug view to prevent crash
- Switch data view text filter from debounced onChange to Enter key
(avoids slow ILIKE queries on every keystroke for large banks)
- Show loading spinner in search icon during filter requests
- Preserve search/tag filters when clicking "Load more"
* chore: sync generated files after rebase
fcntl is a Unix-only module — importing it unconditionally causes an
ImportError on Windows, breaking the entire plugin. Guard the import with a
sys.platform check and fall back to a no-op lock path in
increment_turn_count() so Windows users get correct behaviour without
crashing.
Adds a proper 'none' provider option so users can run Hindsight as a
chunk store with semantic search but without any LLM dependency, replacing
the hacky workaround of setting provider to 'mock'.
When HINDSIGHT_API_LLM_PROVIDER=none:
- Retain automatically uses chunks mode (no fact extraction)
- Recall works normally (semantic search, BM25, graph retrieval)
- Reflect returns HTTP 400 with clear error message
- Consolidation/observations are disabled
- Mental model refresh returns HTTP 400
- No API key required
* feat(reflect): make source facts in search_observations configurable
The recent fix (#669) hardcoded include_source_facts=False in
search_observations to prevent context overflow. This makes it
configurable via HINDSIGHT_API_REFLECT_SOURCE_FACTS_MAX_TOKENS
(env/tenant/bank), defaulting to -1 (disabled).
- -1: source facts disabled (current behavior, default)
- 0: source facts enabled with no token limit
- >0: source facts enabled with a token budget
* docs: add reflect_source_facts_max_tokens to configuration reference
* fix: update configurable fields count in tests and regenerate docs skill
Claude Code's plugin installer does not merge hooks.json into settings.json
automatically. This adds a setup script and skill that users can run once
after installing the plugin to register the hooks manually.
* feat(hermes): file-based config + updated docs
Replace the old dataclass/configure() singleton with a plain dict
config loaded from ~/.hindsight/hermes.json — same field names and
conventions as the openclaw and claude-code integrations.
Loading order: defaults → config file → env var overrides.
- config.py: rewritten with load_config() returning a plain dict,
DEFAULTS matching openclaw/claude-code fields, ENV_OVERRIDES with
typed casting
- tools.py: register() uses load_config() instead of raw env vars
- __init__.py: clean exports (removed configure/get_config/reset_config)
- README.md: full rewrite with config file examples, tables by category
- docs/hermes.md: full rewrite with quick start, architecture, all
config tables, gateway section, troubleshooting
- tests: updated for new config pattern, 46 tests pass
* ci: add test job for hermes integration
* chore: regenerate docs skill for hermes integration
Add a detect-changes job using dorny/paths-filter to determine which
parts of the monorepo changed, then gate each CI job with appropriate
conditions. This avoids running all ~30 jobs for docs-only or
integration-only changes.
Key behaviors:
- Docs/README-only changes only run build-docs and test-doc-examples
- Integration package changes only run their specific test job
- Client SDK changes only run their build/test + dependent jobs
- Core API changes run all API-dependent jobs
- CI config changes (.github/**) run everything as a safety net
- workflow_dispatch (manual) always runs everything
- verify-generated-files always runs unconditionally
* feat(embed): add programmatic UI (control plane) management
Add ability to start/stop the web UI from hindsight-embed, with
configurable port (default: daemon_port + 10000) and hostname
(default: 0.0.0.0). Uses npx to run the published control plane
package, or node directly in dev mode.
New CLI commands:
hindsight-embed ui start [--port PORT] [--hostname HOST]
hindsight-embed ui stop [--port PORT]
hindsight-embed ui status [--port PORT]
hindsight-embed ui logs [-f] [-n N]
New programmatic API:
daemon_client.start_ui(profile, ui_port, hostname)
daemon_client.stop_ui(profile, ui_port)
daemon_client.is_ui_running(profile, ui_port)
daemon_client.get_ui_url(profile, ui_port)
* feat(embed): expose UI management on HindsightEmbedded
Add start_ui(), stop_ui(), is_ui_running(), and ui_url property
to HindsightEmbedded so the UI can be started programmatically:
client = HindsightEmbedded(profile="myapp", ...)
client.start_ui() # starts daemon + UI
print(client.ui_url)
* feat: add LiteLLM LLM provider for Bedrock and 100+ providers
Add a new `litellm` LLM provider that uses the LiteLLM SDK for chat
completions and tool calling, enabling AWS Bedrock and 100+ other
providers for Hindsight's core engine (retain, recall, reflect).
- New LiteLLMLLM provider in engine/providers/litellm_llm.py
- Registered in factory, valid providers list, and no-api-key set
- Refactored API key validation to use requires_api_key() helper
- Added boto3 dependency for Bedrock auth
- Updated docs: configuration, models, monitoring, providers grid
* feat: add bedrock as first-class LLM provider alias
Add `bedrock` as a dedicated provider name that auto-prepends the
`bedrock/` prefix to model names and delegates to LiteLLMLLM under
the hood. This makes Bedrock support more discoverable — users set
`HINDSIGHT_API_LLM_PROVIDER=bedrock` with plain Bedrock model IDs.
* test: add Bedrock to CI provider tests
- Add bedrock/us.amazon.nova-lite-v1:0 to MODEL_MATRIX in test_llm_provider.py
- Add AWS credential check in should_skip_provider()
- Pass AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION_NAME secrets to test-api job
- Update default bedrock model to amazon.nova-2-lite-v1:0
* fix: regenerate docs skill files and bump memory test timeout
- Regenerate skills/hindsight-docs references after docs changes
- Bump test_llm_provider_memory_operations timeout to 600s for slower
providers like Bedrock via LiteLLM
* test: skip bedrock lite models in memory operations test
Nova Lite has a 10K output token limit which is too low for fact
extraction (requires 64K). The api_methods test (completion, tools,
structured output) already validates the provider works correctly.
* test: use Nova Pro for bedrock CI tests to cover full memory pipeline
Nova Lite only supports 10K output tokens, too low for fact extraction.
Switch to Nova Pro which supports the full 64K output needed for
retain/reflect operations. This ensures bedrock is tested on all
Hindsight functionalities, not just basic API methods.
* test: switch bedrock CI to Nova 2 Lite (supports 64K output tokens)
Nova v1 models (Pro, Lite) have a 10K output token limit which is
too low for fact extraction. Nova 2 Lite supports 64K+ output tokens,
enabling full memory pipeline testing (retain + reflect).
MCP tool bridges sometimes serialize JSON arrays as strings during
transport, e.g. '["a", "b"]' arrives as the literal string '["a", "b"]'
instead of a native JSON array. This causes Pydantic to reject the
input with a validation error.
Add defensive coercion at two layers:
1. HTTP API (http.py): Pydantic field_validator on MemoryItem.tags
with mode="before" that parses JSON strings back into lists.
2. MCP tools (mcp_tools.py): Same coercion in build_content_dict
before tags reach the Pydantic model.
A plain non-JSON string is wrapped in a single-element list.
Correctly-formatted input is passed through unchanged.
Co-authored-by: Philipp <philipp@Philipps-MacBook-Pro.local>
Expose the named retain strategy on the MCP retain tool, matching the
HTTP API's per-item strategy support. This allows MCP clients (Claude
Code, Claude Desktop, etc.) to specify extraction behavior per memory:
strategy: "exact" → verbatim storage, no LLM processing
strategy: "verbose" → detailed extraction
strategy: "concise" → default compressed extraction
Strategies are defined in bank config under retain_strategies.
Unknown strategy names are logged and ignored (bank default applies).
Changes:
- Add strategy param to both retain function signatures (with/without bank_id)
- Add strategy to build_content_dict
- Strategy is set in the content dict, which the engine already handles per-item
Co-authored-by: Philipp <philipp@Philipps-MacBook-Pro.local>
Tool handlers and lifecycle hooks now use the native async client API
(aretain, arecall, areflect, acreate_bank) instead of sync wrappers
that call loop.run_until_complete(), which deadlocks in async contexts
like Discord/Telegram gateways.
* fix: return metadata in recall responses (#674)
Metadata stored during retain was never retrieved during recall.
Add metadata to all SQL SELECT queries, the RetrievalResult dataclass,
ScoredResult.to_dict(), and MemoryFact construction in the recall pipeline.
* test: add metadata round-trip test for retain→recall
Replace placeholder metadata test with one that actually passes
metadata via retain_batch_async and asserts it is returned on recall.
* fix: parse metadata JSON string from database in MemoryFact
asyncpg may return JSONB columns as strings. Add a field_validator
to MemoryFact.metadata to handle JSON string deserialization.
* security: exclude litellm 1.82.8 (supply chain compromise)
litellm 1.82.8 on PyPI contains a malicious .pth file that
automatically steals credentials on Python startup (no import needed).
See: https://github.com/BerriAI/litellm/issues/24512
Our Docker images ship 1.82.6 and are unaffected, but the open version
constraints (>=1.0.0, >=1.40.0) would allow resolving to 1.82.8 on
fresh installs or lockfile refreshes.
* security: cap litellm at <=1.82.6 (1.82.7 also compromised)
* chore: regenerate uv.lock and openapi spec
* fix: update test to match claude-haiku-4-5 default model name and regenerate docs skill
* chore: fix ruff formatting in generate_changelog.py
* Add blog post: Adding Long-Term Memory to LangGraph and LangChain Agents
* blog: update langgraph post date to 2026-03-24 and add cover image
* blog: fix claude-code-telegram filename to match frontmatter date (2026-03-25)
* blog: set claude-code-telegram date to 2026-03-23
* blog: fix date timezone offset by adding T12:00 to all post dates
* ci: trigger fresh CI run
* blog: fix broken docs link (routeBasePath is /)
* feat: add Strands Agents SDK integration with Hindsight memory tools
* fix: add strands docs to versioned docs so build link check passes
* fix(strands): run hindsight client calls in thread pool to avoid event loop conflict with Strands
* feat(openclaw): remove hardcoded default models, rely on Hindsight API defaults
* feat(claude-code): remove hardcoded default models, rely on Hindsight API defaults
* feat(claude-code,docs): remove hardcoded default models from claude-code integration and docs
* feat: use claude-haiku-4-5 as default Anthropic model
* docs: add 0.4.20 release blog post and changelog
Add release notes blog post covering Claude Code integration, LangGraph
integration, NemoClaw integration, independent integration versioning,
and reflect improvements. Auto-generated changelog entry included.
* docs: add 0.4.20 release blog cover image
search_observations in the reflect agent hardcoded include_source_facts=True
with max_source_facts_tokens=-1 (unlimited). For banks with many observations
backed by thousands of facts, a single tool call could produce 300K+ tokens,
exceeding the default 100K context budget and causing forced synthesis with
an empty 'Retrieved Data' section.
The reflect agent synthesizes from observations, not raw backing facts.
Disable source facts to keep payloads proportional to observation count
(~6K vs ~310K in the reporter's case).
The consolidation path already has configurable source fact limits (PR #509,
v0.4.17). The reflect path was not updated.
Fixes#668
Co-authored-by: Kagura Chen <daniyuu19@sjtu.edu.cn>
Daemon cold start takes ~25s but hooks have short timeouts, causing
retain to time out on first use. Fix by firing daemon startup as a
detached background process in SessionStart so it warms up before the
first recall/retain hook fires.
Also bumps the daemon start timeout in _ensure_daemon_running from 10s
to 30s as a fallback for when retain fires before pre-start completes.
* fix(entity_resolver): prevent _pending_stats/_pending_cooccurrences memory leak
Add discard_pending_stats() to EntityResolver to clean up both pending dicts
for the current task key. Call it at the start of each _run_db_work attempt so
that exceptions between accumulation and flush_pending_stats() — including
deadlock retries — never leave stale entries keyed by recycled task IDs.
Fixes#660
* test(entity_resolver): add unit tests for discard_pending_stats()
Covers: clears both dicts for current task, is idempotent when empty,
and does not touch entries belonging to other task keys.
No database required — purely in-memory logic.
* doc: add Claude Code + Telegram + Hindsight blog post
* doc: add fabioscarsi to blog authors
* doc: update fabioscarsi title to Contributor
* doc: remove horizontal rule dividers from blog post
* doc: update cover image and add image frontmatter for claude-code-telegram blog post
* doc: remove horizontal rule dividers
* doc: align Hindsight setup steps with PR #661 README
* fix: move marketplace.json to repo root and update source path
* doc: add Claude Code integration page, sidebar, and integrations hub entry
* doc: update versioned docs to 0.4.19
---------
Co-authored-by: Ben <ben.bartholomew@vectorize.io>
* fix(claude-code): fix plugin installation and release workflow
- Fix plugin.json author field (string → object) to pass claude plugin validate
- Add hindsight-integrations/.claude-plugin/marketplace.json so users can install
via: claude plugin marketplace add vectorize-io/hindsight --sparse hindsight-integrations
- Update README and install.sh with correct two-command install flow
- Fix release-integration.yml: add explicit package.json check for typescript type
and add plugin type for integrations with neither pyproject.toml nor package.json
(prevents claude-code from incorrectly falling into the typescript build path)
- Add CHANGELOG.md for the claude-code integration
* remove install.sh — users install via claude plugin commands directly
* test(claude-code): add 116 unit tests for plugin hooks and lib modules
* feat(claude-code): user settings.json at CLAUDE_PLUGIN_DATA for stable config
Plugin now checks CLAUDE_PLUGIN_DATA/settings.json after the versioned
plugin default, giving users a path that persists across updates:
~/.claude/plugins/data/hindsight-memory-hindsight/settings.json
Loading order: defaults → plugin settings.json → user settings.json → env vars
* fix(claude-code): use ~/.hindsight/claude-code.json for user config
Matches the ~/.openclaw/openclaw.json convention. Removes the confusing
CLAUDE_PLUGIN_DATA path whose name depends on marketplace+plugin identifiers.
* docs(claude-code): add ToS hint for claude-code LLM provider option
* fix(claude-code): set author to Hindsight Team in plugin.json
* ci: add test-claude-code-integration job to run plugin unit tests
* feat: Add Claude Code integration plugin
Complete port of hindsight-openclaw (v0.4.19) adapted to Claude Code's
hook-based plugin architecture. Pure Python stdlib, no external dependencies.
- Auto-recall via UserPromptSubmit hook (additionalContext injection)
- Auto-retain via async Stop hook (chunked retention with sliding window)
- Daemon management (auto-start/stop hindsight-embed via uvx)
- Dynamic bank IDs with per-agent/project/channel/user granularity
- All 34 configuration options with env var overrides
- File-based state persistence with fcntl locking
- Graceful degradation on all error paths
Works with Claude Code Channels (Telegram, Discord, Slack) and
interactive sessions.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: Set correct chunked retention defaults (10/2, not 1/0)
retainEveryNTurns=10 and retainOverlapTurns=2 are the production-tested
values — every 10 turns, retain a 12-turn sliding window. The previous
defaults (1/0) would retain every single turn with no overlap, defeating
the chunked retention design that prevents API bombardment.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: Align recallBudget and daemonIdleTimeout with Openclaw defaults
recallBudget: "low" → "mid" (Openclaw default)
daemonIdleTimeout: 300 → 0 (Openclaw default, never auto-stop)
As an official Hindsight integration, defaults should match Openclaw.
Users can optimize locally via settings.json or env vars.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Rename hindsight-openclaw-pro → HindClaw and update description to
reflect the current architecture: server-side Hindsight extensions
(hindclaw-extension on PyPI), Terraform provider for infrastructure
management, and the hindclaw-openclaw gateway plugin.
Link points to https://github.com/mrkhachaturov/hindclaw.
* test: add unit tests for pg_trgm auto-detection and ValidationResult.accept_with() enrichment
Two recent PRs landed without dedicated tests:
- #626/#649 (pg_trgm fallback in EntityResolver): add 5 mocked unit tests
covering the trigram→full fallback, single-check guarantee, and sticky
downgrade behaviour.
- #639 (accept_with() enrichment): add 7 pure unit tests for the factory
method plus 5 integration tests verifying the engine applies enriched
contents (retain) and tags/tag_groups (recall) returned by validators.
Also verifies RecallContext carries tag filter state.
* fix: remove 504 from reflect OpenAPI spec to fix progenitor Rust client build
progenitor-impl-0.11.2 panics with `assertion failed: response_types.len() <= 1`
when an endpoint declares more than one response type. PR #643 added
`responses={504: ...}` to the reflect decorator, which injected a second
response type into the generated OpenAPI spec and broke the Rust client build.
Remove the `responses=` kwarg — the 504 is still raised at runtime via
JSONResponse(status_code=504), it just won't appear in the OpenAPI schema.
Regenerate openapi.json accordingly.
* chore: sync generated files and ruff formatting (lint + docs skill)
On managed PostgreSQL services (e.g. Azure Flexible Server), the pg_trgm
extension may not be available, causing two failures:
1. Migration c1a2b3d4e5f6 crashes on CREATE EXTENSION
2. Even if migration is bypassed, the default 'trigram' entity lookup
strategy uses the % operator which requires pg_trgm, causing retain
background tasks to fail silently
Changes:
- Migration now gracefully skips pg_trgm and index creation if the
extension cannot be loaded
- EntityResolver auto-detects pg_trgm availability on first use and
falls back to 'full' lookup strategy with a warning log
Co-authored-by: coder999999999 <coder999999999@users.noreply.github.com>
Validators can now return enriched data via ValidationResult.accept_with()
instead of only accepting or rejecting operations. The engine applies
returned fields (contents, tags, tag_groups) to the operation parameters.
- Add accept_with() factory to ValidationResult with optional enrichment
fields: contents, tags, tags_match, tag_groups
- Add tags, tags_match, tag_groups to RecallContext so validators can
see current filter state
- Update _validate_operation to return ValidationResult
- Apply enrichment from result at all retain (2 sites) and recall call
sites in MemoryEngine
- Existing validators using accept()/reject() work unchanged
LLM providers like MiniMax wrap JSON responses in markdown code fences
(```json ... ```), causing JSON parse failures and 5-11 retries per
extraction. The existing fence stripping logic was gated to only
"lmstudio" and "ollama" providers (and for Ollama, unreachable due to
the _call_ollama_native redirect).
Changes:
- Extract _strip_code_fences() helper function
- Apply fence stripping to all providers in call() (not just local)
- Add fence stripping safety net to _call_ollama_native()
- Add 10 tests covering bare JSON, fenced JSON, malformed fences,
and real-world MiniMax response format
Fixesvectorize-io/hindsight#645
Co-authored-by: feniix <feniix@desktop>
* fix(recall): reject empty queries with 400 and fix SQL parameter gap causing IndeterminateDatatypeError
When query text contains only punctuation/symbols (no word characters after
normalization), the BM25 arms are skipped but the old code still placed `limit`
at \$3 in the params list. If tags or tag_groups were also set, their params
(\$4+) were referenced in the SQL while \$3 was a gap, causing PostgreSQL to
raise IndeterminateDatatypeError.
Fix the parameter layout so `limit` is only appended to params when tokens are
present (i.e. when BM25 arms actually use LIMIT \$3), and shift tags_param_idx
from 4 to 3 in the no-tokens path.
Also add a field_validator on RecallRequest.query that rejects empty-after-
normalization queries at the API layer with a 400 before they reach the DB.
* refactor: extract tokenize_query helper and reuse in RecallRequest validator
Remove the sys_platform == 'darwin' constraint that prevented
claude-agent-sdk from installing on Linux, breaking the claude-code
provider in Docker containers.
Fixes#640
