Skip to content
Problems.

LlamaIndex upgrades break working setups

Outages and change7 posts from 4 people4d active+33 posts in the last 7 days, 3 the 7 days before (steady)

Posts per day

Posts per day7 posts, Sep 8 to Sep 25

The posts behind it

7, newest first
PostDate
[Bug]: DirectLookaheadAnswerInserter splices answers at stale offsets when a lookahead has multiple [Search()] tags### Bug Description `DirectLookaheadAnswerInserter.insert` walks `zip(query_tasks, answers)` and rebuilds the response string for each task: The `start_idx`/`end_idx` on each `QueryTask` point into the original response string (that is where `QueryTaskOutputParser` measured...run-llama/llama_indexHarsh23KashyapSep 223 days ago
[Bug]: HTMLNodeParser rips inline tag text out of its sentence and emits it as a separate node### Bug Description `HTMLNodeParser`'s default tag list includes the inline formatting tags `b`, `i`, and `u`. When one of these appears inside a block tag, the parser drops the inline tag's text from the surrounding sentence and emits that text as its own node after the block....run-llama/llama_indexHarsh23KashyapSep 223 days ago
[Bug]: BaseWorkflowAgent.init_run raises IndexError when chat_history contains no user messages### Bug Description In `llama_index.core.agent.workflow.base_agent.BaseWorkflowAgent.init_run`, when initializing an agent workflow with a `chat_history` that contains non-system messages but no `user` messages (e.g., resuming from an initial assistant greeting or customer...run-llama/llama_indextruecallerabrehamSep 214 days ago
[Bug]: RichPromptTemplate crashes on embedded audio/video files and corrupts embedded documents...URL, relative URL without a base`. Any chat template embedding an audio or video file is unusable. - **Document files are silently corrupted.** `DocumentBlock.url` is a plain string, so the base64 blob "succeeds" and every downstream consumer receives a document block whose...run-llama/llama_indexHarsh23KashyapSep 169 days ago
Ingestion cache contract: stale node identity vs content-stable reuse after regenerated chunk IDs## Context PR #23003 fixes the document-level identity collision from #23002, but its follow-up discussion exposed a broader cache-contract choice that appears worth deciding separately rather than burying in that focused fix. The current generic transformation cache stores and...run-llama/llama_indexNakagawa-masterSep 169 days ago
[Bug]: SimplePropertyGraphStore.get_rel_map returns duplicate triplets when the graph has cycles### Bug Description `SimplePropertyGraphStore.get_rel_map` walks the graph depth by depth and tracks visited triplets in `seen_triplets`, but the seen-set starts empty and is only updated with the *next* depth's triplets: Any path that loops back to an already-visited depth - a...run-llama/llama_indexHarsh23KashyapSep 169 days ago
McpToolSpec exposes dynamic MCP resource templates as zero-argument read_resource tools...a URI template such as: The generated FunctionTool has no arguments and therefore has no way to bind `owner`/`name` before calling `read_resource()`. It attempts to read the template URI literally. ## Minimal Reproduction Shape A fake MCP client is sufficient: Then: The...run-llama/llama_indexanupamking01Sep 82 weeks ago

Companies and products named

Company or productPosts naming it
LlamaIndex7
Model Context Protocol1
About this problem

Evidence

7 posts from 4 people in 1 place, about 2 a week over 15 days. Mostly on run-llama/llama_index. Tools named alongside: Model Context Protocol.

Frustration Frustration 2 of 3· Seen on GitHub issues

How it was grouped

Posts that state a pain and match the "breaking-changes" rule about LlamaIndex. First post Sep 8, 2026, latest Sep 22, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method

History

  • 2026-09-25 Added: LlamaIndex upgrades break working setups (7 posts)

Rising problems by email

Mondays: the problems in data, tech and AI that grew fastest that week.

Double opt-in. Unsubscribe any time.