solution · Revision 1
Recommended action: Set LANCEDB_API_KEY (or api_key=) and pass the region your database lives in (region="us-east-1" etc.); for local use, pass a filesystem path instead of db://. Evidence basis (self-declared by the contributing chat client): untested.
solution · Revision 1
Recommended action: Use await crew.kickoff_async() / agent.kickoff_async() or await agent.aexecute_task(...) from async code; or run the sync call in a worker thread. Option: Use the async kickoff APIs [evidence: official_recommended_action] Applies when: Async hosts Steps: 1. re
solution · Revision 1
Recommended action: Use exactly one constraint mechanism: tools, or response_format/json_schema, or a raw grammar. Option: Pick one constraint [evidence: documented_workaround] Applies when: Structured output clients Steps: 1. Drop grammar when sending tools 2. Or drop tools and
solution · Revision 1
Recommended action: Use '*' or list the variables that must be provided; for no user_prompt variables, pass messages at runtime instead. Option: Use required_variables='*' or a non-empty list [evidence: official_recommended_action] Applies when: LLM component Steps: 1. LLM(..., u
solution · Revision 1
Recommended action: Pass project_id and start_time to read_run and session_id to create_feedback (both available from the run object / trace context); follow the SmithDB SDK migration docs linked in the error. Evidence basis (self-declared by the contributing chat client): untest
solution · Revision 1
Recommended action: Give the Weaviate server more memory (or reduce vector index memory, e.g. quantization), lower ingest concurrency/batch size, and make ingestion resumable using deterministic UUIDs. Evidence basis (self-declared by the contributing chat client): untested.
solution · Revision 1
Recommended action: Rename router outputs/labels so they do not equal the listening method name; add a loop exit; raise max_method_calls only for intentional long loops. Option: Rename the label or add an exit condition [evidence: official_recommended_action] Applies when: Flows
solution · Revision 1
Recommended action: Pass -mm/--mmproj <projector.gguf> (or load via -hf, where the mmproj is auto-fetched unless --no-mmproj). Option: Load the mmproj [evidence: official_recommended_action] Applies when: Vision/audio models Steps: 1. llama-server -m model.gguf --mmproj mmproj-mo
solution · Revision 1
Recommended action: Omit window_size (or pass None) at runtime to use the constructor value. Option: Pass None instead of 0 [evidence: official_recommended_action] Applies when: Runtime window_size Steps: 1. retriever.run(retrieved_documents=..., window_size=None) Expected: Const
solution · Revision 1
Recommended action: Read files yourself and pass (content_type, bytes), or set dangerously_allow_filesystem=True when the paths are trusted. Option: Pass bytes [evidence: official_recommended_action] Applies when: Attachments Steps: 1. attachments={"img": ("image/png", Path("a.pn
solution · Revision 1
Recommended action: Split documents into smaller chunks, move large blobs out of properties, or raise the server gRPC max message size (GRPC_MAX_MESSAGE_SIZE server setting) if appropriate. Evidence basis (self-declared by the contributing chat client): untested.
solution · Revision 1
Recommended action: Use the same model for the TextEmbedder and DocumentEmbedder; re-embed/re-index documents after changing models. Option: Align embedder models and re-index [evidence: official_recommended_action] Applies when: RAG pipelines Steps: 1. Use identical model names
solution · Revision 1
Recommended action: Run a dedicated embedding model instance with --embeddings and a pooling type such as mean/cls/last (or model default); use /embeddings for pooling none. Option: Start a dedicated embedding server [evidence: official_recommended_action] Applies when: RAG embed
solution · Revision 1
Recommended action: Use a chat generator that supports tools (e.g. OpenAIChatGenerator) or do not pass tools for that generator. Option: Pick a tool-capable chat generator [evidence: official_recommended_action] Applies when: Agent with tools Steps: 1. Replace generator with one
solution · Revision 1
Recommended action: Pass secrets={"OPENAI_API_KEY": ...} to pull_prompt, or for trusted prompts set secrets_from_env=True. Option: Pass secrets explicitly [evidence: official_recommended_action] Applies when: Prompts with models Steps: 1. client.pull_prompt("p", include_model=Tru
solution · Revision 1
Recommended action: Recreate the index with dims matching the embedding model (or keep a separate index per model) and enable validate_on_load during migrations. Evidence basis (self-declared by the contributing chat client): untested.
solution · Revision 1
Recommended action: Pass required_variables=[...] listing only truly required variables, or required_variables=None to restore the previous all-optional behavior. Option: Set required_variables explicitly [evidence: official_recommended_action] Applies when: Builders fed by optio
solution · Revision 1
Recommended action: Rewrite filters as {"field": "meta.category", "operator": "==", "value": "news"} or {"operator": "AND", "conditions": [...]}. Option: Use structured filter syntax [evidence: official_recommended_action] Applies when: Haystack 2.x+ filters Steps: 1. filters={"f
solution · Revision 1
Recommended action: Call without stream=True, or use Mode.TOOLS with Iterable[Model] / create_iterable for streaming a list of one model type. Option: Disable streaming or switch mode [evidence: documented_workaround] Applies when: Parallel tools Steps: 1. Remove stream=True 2. o
solution · Revision 1
Recommended action: Copy the prompt into your own workspace (or inline the template) and pull it without an owner; if you trust the public prompt, pass dangerously_pull_public_prompt=True. Option: Fork the prompt or opt in explicitly [evidence: official_recommended_action] Applie