* fix(core): raise LLMEmptyResponseError for empty Anthropic responses
Fixes LET-7679: Opus 4.6 occasionally returns empty responses (no content
and no tool calls), causing silent failures with stop_reason=end_turn.
Changes:
- Add LLMEmptyResponseError class (subclass of LLMServerError)
- Raise error in anthropic_client for empty non-streaming responses
- Raise error in anthropic_streaming_interface for empty streaming responses
- Pass through LLMError instances in handle_llm_error to preserve specific types
- Add test for empty streaming response detection
This allows clients (letta-code) to catch this specific error and implement
retry logic with cache-busting modifications.
🤖 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* fix(core): set invalid_llm_response stop reason for empty responses
Catch LLMEmptyResponseError specifically and set stop_reason to
invalid_llm_response instead of llm_api_error. This allows clients
to distinguish empty responses from transient API errors.
🤖 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
---------
Co-authored-by: Letta <noreply@letta.com>
* feat(core): add gpt-5.3-codex model support
Add OpenAI gpt-5.3-codex model: context window overrides, model pricing
and capabilities, none-reasoning-effort support, and test config.
🐾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* just stage-api && just publish-api
---------
Co-authored-by: Letta <noreply@letta.com>
* fix(core): prevent event loop saturation from ClickHouse and socket trace writes
Two issues were causing the event loop watchdog to fire and liveness probes
to fail under load:
1. LLMTraceWriter held an asyncio.Lock across each ClickHouse write, and
wait_for_async_insert=1 meant each write held that lock for ~1s. Under high
request volume, N background tasks all queued for the lock simultaneously,
saturating the event loop with task management overhead. Fix: switch to
wait_for_async_insert=0 (ClickHouse async_insert handles server-side batching
— no acknowledgment wait needed) and remove the lock (clickhouse_connect uses
a thread-safe connection pool). The sync insert still runs in asyncio.to_thread
so it never blocks the event loop. No traces are dropped.
2. SocketProviderTraceBackend spawned one OS thread per trace with a 60s socket
timeout. During crouton restarts, threads accumulated blocking on sock.sendall
for up to 3 minutes each (3 retries x 60s). Fix: reduce socket timeout from
60s to 5s — the socket is local (Unix socket), so 5s is already generous, and
fast failure lets retries resolve before threads pile up.
Root cause analysis: event_loop_watchdog.py was detecting saturation (lag >2s)
every ~60s on gke-letta-default-pool-c6915745-fmq6 via thread dumps. The
saturated event loop caused k8s liveness probes to time out, triggering restarts.
* chore(core): sync socket backend with main and document ClickHouse thread safety
- Remove `limit` from YAML frontmatter in `serialize_block()` and
`merge_frontmatter_with_body()` (deprecated for git-base memory)
- Remove `limit` from `_render_memory_blocks_git()` in-context rendering
- Existing frontmatter with `limit` is automatically cleaned up on next write
- Parsing still accepts `limit` from frontmatter for backward compatibility
- Increase `CORE_MEMORY_BLOCK_CHAR_LIMIT` from 20,000 to 100,000
- Update integration tests to assert `limit` is not in frontmatter
Fixes#9536
Co-authored-by: letta-code <248085862+letta-code@users.noreply.github.com>
Co-authored-by: Sarah Wooders <sarahwooders@users.noreply.github.com>
Co-authored-by: Sarah Wooders <sarahwooders@gmail.com>
* feat: recompile system message on new conversation creation
When a new conversation is created, the system prompt is now recompiled
with the latest memory block values and metadata instead of starting
with no messages. This ensures each conversation captures the current
agent state at creation time.
- Add _initialize_conversation_system_message to ConversationManager
- Compile fresh system message using PromptGenerator during conversation creation
- Add integration tests for the full workflow (modify memory → new conversation
gets updated system message)
- Update existing test expectations for non-empty conversation messages
Fixes#9507
Co-authored-by: Sarah Wooders <sarahwooders@users.noreply.github.com>
* refactor: deduplicate system message compilation into ConversationManager
Consolidate the duplicate system message compilation logic into a single
shared method `compile_and_save_system_message_for_conversation` on
ConversationManager. This method accepts optional pre-loaded agent_state
and message_manager to avoid redundant DB loads when callers already have
them.
- Renamed _initialize_conversation_system_message → compile_and_save_system_message_for_conversation (public, reusable)
- Added optional agent_state and message_manager params
- Replaced 40-line duplicate in helpers.py with a 7-line call to the shared method
- Method returns the persisted system message for caller use
Co-authored-by: Sarah Wooders <sarahwooders@users.noreply.github.com>
---------
Co-authored-by: letta-code <248085862+letta-code@users.noreply.github.com>
Co-authored-by: Sarah Wooders <sarahwooders@users.noreply.github.com>
* add self compaction method with proper caching (pass in tools, don't refresh sys prompt beforehand) + sliding fallback
* updated prompts for self compaction
* add tests for self, self_sliding_window modes and w/o refresh messages before compaction
* add cache logging to summarization
* better handling to prevent agent from continuing convo on self modes
* if mode changes via summarize endpoint, will use default prompt for the new mode
---------
Co-authored-by: Amy Guan <amy@letta.com>
fix(core): ensure otid exists when flushing buffered anthropic tool chunks
Anthropic TOOL_USE buffering can emit buffered tool_call/approval chunks on content block stop before otid is assigned in the normal inner_thoughts_complete path. Ensure flush-time chunks get a deterministic otid so streaming clients can reliably correlate deltas.
👾 Generated with [Letta Code](https://letta.com)
Co-authored-by: Letta <noreply@letta.com>
* feat: add order_by and order params to /v1/conversations list endpoint [LET-7628]
Added sorting support to the conversations list endpoint, matching the pattern from /v1/agents.
**API Changes:**
- Added `order` query param: "asc" or "desc" (default: "desc")
- Added `order_by` query param: "created_at" or "last_run_completion" (default: "created_at")
**Implementation:**
**created_at ordering:**
- Simple ORDER BY on ConversationModel.created_at
- No join required, fast query
- Nulls not applicable (created_at always set)
**last_run_completion ordering:**
- LEFT JOIN with runs table using subquery
- Subquery: MAX(completed_at) grouped by conversation_id
- Uses OUTER JOIN so conversations with no runs are included
- Nulls last ordering (conversations with no runs go to end)
- Index on runs.conversation_id ensures performant join
**Pagination:**
- Cursor-based pagination with `after` parameter
- Handles null values correctly for last_run_completion
- For created_at: simple timestamp comparison
- For last_run_completion: complex null-aware cursor logic
**Performance:**
- Existing index: `ix_runs_conversation_id` on runs table
- Subquery with GROUP BY is efficient for this use case
- OUTER JOIN ensures conversations without runs are included
**Follows agents pattern:**
- Same parameter names (order, order_by)
- Same Literal types and defaults
- Converts "asc"/"desc" to ascending boolean internally
🐾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* chore: order
---------
Co-authored-by: Letta <noreply@letta.com>
fix: align ChatGPT OAuth GPT-5 max output token defaults
Update ChatGPT OAuth provider defaults so GPT-5 family models report 128k max output tokens based on current OpenAI model docs, avoiding incorrect 16k values in /v1/models responses.
👾 Generated with [Letta Code](https://letta.com)
Co-authored-by: Letta <noreply@letta.com>
* feat: make agent_id optional in conversations list endpoint [LET-7612]
Allow listing all conversations without filtering by agent_id.
**Router changes (conversations.py):**
- Changed agent_id from required (`...`) to optional (`None`)
- Updated description to clarify behavior
- Updated docstring to reflect optional filtering
**Manager changes (conversation_manager.py):**
- Updated list_conversations signature: agent_id: str → Optional[str]
- Updated docstring to clarify optional behavior
- Summary search query: conditionally adds agent_id filter only if provided
- Default list logic: passes agent_id (can be None) to list_async
**How it works:**
- Without agent_id: returns all conversations for the user's organization
- With agent_id: returns conversations filtered by that agent
- list_async handles None gracefully via **kwargs pattern
**Use case:**
- Cloud UI can list all user conversations across agents
- Still supports filtering by agent_id when needed
🐾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* chore: update logs
* chore: update logs
---------
Co-authored-by: Letta <noreply@letta.com>
When model_settings is sent without max_output_tokens (e.g. only
changing reasoning_effort), the Pydantic default of 4096 was being
applied via _to_legacy_config_params(), silently overwriting the
agent's existing max_tokens.
Use model_fields_set to detect when max_output_tokens was not
explicitly provided and skip overwriting max_tokens in that case.
Only applied to the update path — on create, letting the default
apply is reasonable since there's no pre-existing value.
* fix: migrate Anthropic Haiku test model off retired release
Update Anthropic Haiku references in integration and usage parsing tests to a supported model id so test requests stop failing with 404 model not found errors.
👾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* fix: use canonical Anthropic Haiku handle in tests
Replace dated Anthropic Haiku handle references with the canonical provider handle so handle-based model resolution does not fail in batch and client tests.
👾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
---------
Co-authored-by: Letta <noreply@letta.com>
* load default settings instead of loading from agent for summarizer config
* update tests to allow use of get_llm_config_from_handle
* remove nit comment
---------
Co-authored-by: Amy Guan <amy@letta.com>
The Pydantic validator `set_model_specific_defaults` was checking
`values.get("max_tokens") is None`, which matched both "field not
provided" and "field explicitly set to null". This meant users could
not disable the max output tokens limit for GPT-5/GPT-4.1 models -
the validator would always override null with a default value during
request deserialization.
Changed to `"max_tokens" not in values` so that an explicit
`max_tokens: null` is preserved while still applying defaults when
the field is omitted entirely.
list_llm_models_async was constructing LLMConfig without max_tokens,
causing the GET /models/ endpoint to return null for max_tokens.
Now calls typed_provider.get_default_max_output_tokens() for both
base and BYOK provider paths, matching get_llm_config_from_handle.
Add 3.1 model metadata for Google AI and update Gemini tests/examples to use the new handle.
👾 Generated with [Letta Code](https://letta.com)
Co-authored-by: Letta <noreply@letta.com>
perf(memfs): delta upload — only push new/modified git objects after commit
Instead of re-uploading the entire .git/ directory after every commit,
snapshot file mtimes before the commit and only upload files that are
new or changed. A typical single-block update creates ~5 new objects
(blob, trees, commit, ref) vs re-uploading all ~30.
Full _upload_repo retained for create_repo and other paths that need it.
🤖 Generated with [Letta Code](https://letta.com)
Co-authored-by: Letta <noreply@letta.com>
LlmConfig._to_model_settings() for Anthropic built an AnthropicModelSettings
object without passing effort=self.effort, so GET /agents/{id} never returned
the effort field in model_settings even when it was stored on the agent.
The Letta Code CLI derives the reasoning tier displayed in the status bar
from model_settings.effort (canonical source), so the footer always showed
e.g. "Sonnet 4.6" instead of "Sonnet 4.6 (high)" after a model switch.
👾 Generated with [Letta Code](https://letta.com)
Co-authored-by: Letta <noreply@letta.com>
* fix(core): always create system message even with _init_with_no_messages
When _init_with_no_messages=True (used by agent import flows), the agent
was created with message_ids=None. If subsequent message initialization
failed, this left orphaned agents that crash when context window is
calculated (TypeError on message_ids[1:]).
Now the system message is always generated and persisted, even when
skipping the rest of the initial message sequence. This ensures every
agent has at least message_ids=[system_message_id].
Fixes Datadog issue 773a24ea-eeb3-11f0-8f9f-da7ad0900000
🐾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* fix(core): clean up placeholder messages during import and add test
Delete placeholder system messages after imported messages are
successfully created (not before), so agents retain their safety-net
system message if import fails. Also adds a test verifying that
_init_with_no_messages=True still produces a valid context window.
🐾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* fix(core): add descriptive error for empty message_ids in get_system_message
🐾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
---------
Co-authored-by: Letta <noreply@letta.com>
* fix(core): handle ResponseIncompleteEvent in OpenAI Responses API streaming
When reasoning models (gpt-5.x) exhaust their max_output_tokens budget
on chain-of-thought reasoning, OpenAI emits a ResponseIncompleteEvent
instead of ResponseCompletedEvent. This was previously unhandled, causing
final_response to remain None — which meant get_content() and
get_tool_call_objects() returned empty results, silently dropping the
partial response.
Now ResponseIncompleteEvent is handled identically to
ResponseCompletedEvent (extracting partial content, usage stats, and
token details), with an additional warning log indicating the incomplete
reason.
* fix(core): propagate finish_reason for Responses API incomplete events
- Guard usage extraction against None usage payload in
ResponseIncompleteEvent handler
- Add _finish_reason override to LettaLLMAdapter so streaming adapters
can explicitly set finish_reason without a chat_completions_response
- Map incomplete_details.reason="max_output_tokens" to
finish_reason="length" in SimpleLLMStreamAdapter, matching the Chat
Completions API convention
- This allows the agent loop's _decide_continuation to correctly return
stop_reason="max_tokens_exceeded" instead of "end_turn" when the model
exhausts its output token budget on reasoning
* fix(core): handle empty content parts in incomplete ResponseOutputMessage
When a model hits max_output_tokens after starting a ResponseOutputMessage
but before producing any content parts, the message has content=[]. This
previously raised ValueError("Got 0 content parts, expected 1"). Now it
logs a warning and skips the empty message, allowing reasoning-only
incomplete responses to be processed cleanly.
* fix(core): map all incomplete reasons to finish_reason, not just max_output_tokens
Handle content_filter and any future unknown incomplete reasons from the
Responses API instead of silently leaving finish_reason as None.
* fix(core): add OpenAI prompt cache key and model-gated 24h retention (#9492)
* fix(core): apply OpenAI prompt cache settings to request payloads
Set prompt_cache_key using agent and conversation context on both Responses and Chat Completions request builders, and enable 24h retention only for supported OpenAI models while excluding OpenRouter paths.
👾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* fix(core): prefix prompt cache key with letta tag
Add a `letta:` prefix to generated OpenAI prompt_cache_key values so cache-related entries are easier to identify in provider-side logs and diagnostics.
👾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* add integration test
* skip test
---------
Co-authored-by: Letta <noreply@letta.com>
Co-authored-by: Ari Webb <ari@letta.com>
* fix(core): only set prompt_cache_retention, drop prompt_cache_key
Two issues with the original prompt_cache_key approach:
1. Key exceeded 64-char max (agent-<uuid>:conv-<uuid> = 90 chars)
2. Setting an explicit key disrupted OpenAI's default prefix-hash
routing, dropping cache hit rates from 40-45% to 10-13%
OpenAI's default routing (hash of first ~256 tokens) already provides
good cache affinity since each agent has a unique system prompt.
We only need prompt_cache_retention="24h" for extended retention.
Also fixes:
- Operator precedence bug in _supports_extended_prompt_cache_retention
- Removes incorrect gpt-5.2-codex exclusion (it IS supported per docs)
🐾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
---------
Co-authored-by: Charles Packer <packercharles@gmail.com>
Co-authored-by: Letta <noreply@letta.com>
Adds a diagnostic log at the streaming chokepoint in LettaAgentV3.stream()
to detect when any LettaMessage chunk is yielded without an id or otid field.
This helps trace the root cause of client-side id/otid inconsistencies.
* auto fixes
* auto fix pt2 and transitive deps and undefined var checking locals()
* manual fixes (ignored or letta-code fixed)
* fix circular import
* remove all ignores, add FastAPI rules and Ruff rules
* add ty and precommit
* ruff stuff
* ty check fixes
* ty check fixes pt 2
* error on invalid
* fix(core): apply OpenAI prompt cache settings to request payloads
Set prompt_cache_key using agent and conversation context on both Responses and Chat Completions request builders, and enable 24h retention only for supported OpenAI models while excluding OpenRouter paths.
👾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* fix(core): prefix prompt cache key with letta tag
Add a `letta:` prefix to generated OpenAI prompt_cache_key values so cache-related entries are easier to identify in provider-side logs and diagnostics.
👾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta <noreply@letta.com>
* add integration test
* skip test
---------
Co-authored-by: Letta <noreply@letta.com>
Co-authored-by: Ari Webb <ari@letta.com>
