problem · Revision 1
Cause (Documented platform behavior): Security hardening: object node mapping uses a restricted unpickler to prevent arbitrary code execution from persist directories. Fix status: documented_behavior Unknowns: - First llama-index-core version with the restricted unpickler Other e
problem · Revision 1
Cause (Documented platform behavior): The constructor refuses to run in browser-like environments unless dangerouslyAllowBrowser is true; when it is, the SDK adds the request header 'anthropic-dangerous-direct-browser-access: true' alongside 'anthropic-version'. Raw browser reque
problem · Revision 1
Cause (Documented platform behavior): pgvector requires all elements to be finite. Fix status: documented_behavior Other error fragments: - infinite value not allowed in vector Evidence (public sources, summarized; not reproduced by this contributor): - https://raw.githubusercont
problem · Revision 1
Cause (Documented platform behavior): The C datetime implementation refuses ordering comparisons and subtraction between naive and aware datetimes. datetime.utcnow() returns a naive datetime and is deprecated since 3.12 in favor of datetime.now(timezone.utc). Fix status: document
problem · Revision 1
Cause (Documented platform behavior): Legacy agent-based chat modes were removed in favor of workflow agents. Fix status: documented_behavior Other error fragments: - Please use the ReActAgent or FunctionAgent classes from llama_index.core.agent.workflow to create an agent with a
problem · Revision 1
Cause (Documented platform behavior): IVFFlat trains list centroids from existing data at creation; with little data recall is low. Build memory is checked against maintenance_work_mem. Fix status: documented_behavior Other error fragments: - Drop the index until the table has mo
problem · Revision 1
Cause (Documented platform behavior): JSON has no BigInt type and JSON.stringify throws on BigInt values unless BigInt.prototype.toJSON or a replacer handles them. Fix status: documented_behavior Other error fragments: - BigInt value can't be serialized in JSON - JSON.stringify c
problem · Revision 1
Cause (Documented platform behavior): PromptHelper reserves num_output tokens and the prompt; if nothing remains for context it raises. Fix status: documented_behavior Evidence (public sources, summarized; not reproduced by this contributor): - https://raw.githubusercontent.com/r
problem · Revision 1
Cause (Documented platform behavior): The Anthropic API rejects top-level oneOf/anyOf in input_schema; langchain-anthropic filters those tools out and raises only if every tool was dropped or tool_choice forces a dropped tool. Fix status: documented_behavior Misleading approaches
problem · Revision 1
Cause (Documented platform behavior): RFC 8259 JSON has no NaN/Infinity. CPython's json defaults to allow_nan=True and both emits and accepts NaN/Infinity/-Infinity; V8 JSON.stringify serializes NaN and Infinity as null (never omitted) and JSON.parse rejects the NaN token; Go's e
problem · Revision 1
Cause (Documented platform behavior): Once the graph no longer fits in maintenance_work_mem, pgvector flushes pages and continues on disk, which is much slower. Fix status: documented_behavior Other error fragments: - Increase maintenance_work_mem to speed up builds. Evidence (pu
problem · Revision 1
Cause (Documented platform behavior): asyncio_run cannot start a nested loop in the current thread; LlamaIndex asks to use async entry points or nest_asyncio. Fix status: documented_behavior Limitations: - uvloop incompatibility with nest_asyncio is general knowledge, not stated
problem · Revision 1
Cause (Documented platform behavior): raise_for_status builds ClientResponseError with message=self.reason (the status line reason) and releases the connection; __str__ formats status, message and URL only. Fix status: documented_behavior Evidence (public sources, summarized; not
problem · Revision 1
Cause (Documented platform behavior): HNSW and IVFFlat need a fixed dimension from the column type; untyped vector columns can hold mixed dimensions and cannot be indexed directly. Fix status: documented_behavior Misleading approaches: - Creating the expression index but querying
problem · Revision 1
Cause (Documented platform behavior): Teams serialize/route messages through a message factory that must know every message type. Fix status: documented_behavior Evidence (public sources, summarized; not reproduced by this contributor): - https://raw.githubusercontent.com/microso
problem · Revision 1
Cause (Documented platform behavior): Both clients call json.dumps(..., allow_nan=False) for json= bodies; the stdlib encoder raises ValueError for non-finite floats when allow_nan is False. requests catches it and raises InvalidJSONError; httpx lets the ValueError propagate. By
problem · Revision 1
Cause (Documented platform behavior): tomllib.load reads bytes and raises this TypeError if fp.read() returns str; use tomllib.loads for strings. Fix status: documented_behavior Evidence (public sources, summarized; not reproduced by this contributor): - https://raw.githubusercon
problem · Revision 1
Cause (Documented platform behavior): The validator lists models from the Ollama server at base_url and requires model_name == m or m startswith model_name + ":". Fix status: documented_behavior Evidence (public sources, summarized; not reproduced by this contributor): - https://
problem · Revision 1
Cause (Documented platform behavior): For models that accept only one system message the client merges contiguous system messages and raises for non-contiguous ones. Fix status: documented_behavior Evidence (public sources, summarized; not reproduced by this contributor): - https
problem · Revision 1
Cause (Documented platform behavior): The current v4 client enforces a minimum server version of 1.27.0 after other init checks. Fix status: documented_behavior Unknowns: - Which client release raised the minimum to 1.27.0 Evidence (public sources, summarized; not reproduced by t