> ## Documentation Index > Fetch the complete documentation index at: https://aidocs.ethanbragdon.icu/llms.txt > Use this file to discover all available pages before exploring further. # Files & Knowledge Troubleshooting > Troubleshoot Starfire AI uploads, extraction, indexing, stale sources, permissions, retrieval, and repository-backed Knowledge. # Files & Knowledge troubleshooting Use this guide when a file uploads but cannot be used, a Knowledge source never becomes ready, or retrieval misses expected content. ## Upload failed Check: * file size and type * account/organization upload policy * storage limit * active session * current platform health ## Upload succeeded but content is unavailable The file can exist before extraction or indexing is complete. Check processing state. ## Source is stuck processing Determine which stage is stuck: * extraction * chunking * embeddings * indexing If many unrelated sources are stuck at the same stage, inspect worker/provider health. ## Source is `FAILED` Inspect source-specific error metadata when available. Fix unsupported/corrupt input, integration access, provider problems, or storage issues before retrying. ## Source is ready but query misses it Check: 1. active Project/organization context 2. requester permission 3. direct distinctive query 4. retrieval candidates 5. reranking/filtering 6. context/result limits 7. model interpretation ## Repository is stale Refresh the connected repository/revision after major changes. Confirm Starfire still has integration access and is indexing the intended branch. ## Duplicate sources Avoid repeated re-upload as a troubleshooting technique. Duplicates can create confusing retrieval results while leaving the real ingestion problem unresolved. ## Citation points to wrong material Treat that as a retrieval/citation-verification issue. A source existing in the run is not enough; the cited passage must support the claim. Follow the full ingestion-to-model diagnostic chain.