fix(deps): update dependency llama-index to v0.12.22

[renovate]

This PR contains the following updates:

Package	Change	Age	Adoption	Passing	Confidence
llama-index (source)	0.12.14 -> 0.12.22

---

Release Notes

run-llama/llama_index (llama-index)

v0.12.22

Compare Source

• fix agentworkflow tool call tracking on final response (#​17968)

v0.12.21

Compare Source

• fix: remove warnings from workflow tests (#​17943)
• fix: take step workers into account when running a workflow step-wise (#​17942)
• feat: auto-detect custom start and stop events in workflow classes (#​17865)
• Feature/remove retriever tool template override (#​17909)
• only modify delta if 'Answer:' was actually detected (#​17901)
• Fix CitationQueryEngine init function for response_synthesizer (#​17897)
• fix ChatSummaryMemoryBuffer._summarize_oldest_chat_history (#​17845)
• fix: make base64 detection more robust across the board (#​17930)
• fix: stepwise execution breaks when steps do async work (#​17914)
• safer workflow cancel + fix restored context bug (#​17938)

v0.12.20

Compare Source

• feat: Attach more metadata to retrieved image node (#​17868)
• fix: FunctionAgent and ReActAgent error out when LLM response contains no chat messages (#​17884)
• fix: Convert Embedding Functions to Async (#​17888)

v0.12.19

Compare Source

• Added sync/async Function tool callback (#​16637)
• Tweaked regex for template vars

v0.12.18

Compare Source

• improve prompt helper multimodal support (#​17831)
• Retain return type from @dispatcher.span (#​17817)
• Silence nltk downloader (#​17816)
• Keep a reference to asyncio tasks in astream_chat() (#​17812)

v0.12.17

Compare Source

• Added support for AudioBlocks in chat messages (OpenAI only for now) (#​17780)
• Fix: SimpleDirectoryReader metadata handles timezones inconsistently (#​17724)

v0.12.16

Compare Source

• Be more lenient with leading whitespaces emitted by some models when doing ReAct (#​17701)
• Fix user_msg vs chat_history AgentWorkflow inputs (#​17690)

v0.12.15

Compare Source

• Add error_on_tool_error param to FunctionCallingLLM.predict_and_call (#​17663)
• Get tool description from pydantic field (#​17679)
• fix: make ctx._events_buffer json-serializable (#​17676)
• feat: allow to exclude empty file simple directory reader (#​17656)
• improve markdown llm output parsing (#​17577)
• small typo fix in the default plan refine prompt (#​17644)

---

Configuration

📅 Schedule: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined).

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.