# What changed in nv-ingest-client

> 37 changes in nv-ingest-client 26 from its official release notes, upgrade guides and advisories: 6 breaking, 17 new APIs and options. Each one cited. Sources checked 2026-10-31. HTML: https://unstale.sh/pypi/nv-ingest-client/

To see which of these apply to a repository, run `npx @unstale/cli init` in it.

## nv-ingest-client 26

### Breaking changes

- Package installs no longer support macOS Intel x86_64; Apple Silicon macOS, Windows x64, and Linux remain supported. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Helm replaces separate page-elements and table-structure NIM images with the combined object-detection image. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Legacy nv-ingest and compatibility pipeline CLI code paths are removed; use retriever ingest and the graph stage registry. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- The minimum Ray version is raised to >=2.56.1 for CVE remediation. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- First-class ingestion command and canonical replacement for the retired compatibility pipeline command. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- An unacknowledged VectorDB write now fails instead of reporting completed with a positive row count. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)

### New APIs and options

- Adds an MCP tool for agentic retrieval when agentic service functionality is enabled. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Accepts a validated custom prompt for generic text generation. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Adds documented Markdown, JSON, and shell text inputs, plus inline text ingestion support. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Adds an optional embedding checkpoint for local Hugging Face inference and hosted or self-hosted embedding NIM use. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Adds endpoint wiring in service extraction workers, with distinct hosted and self-hosted HTTP contracts. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Enables agentic retrieval on the service query endpoint when set to true. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- First-class query command with NIM URL flags, batch tuning, and LanceDB overwrite/append controls. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Runs an LLM-driven ReAct retrieval loop over the same LanceDB table as one-pass retrieval. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Selects a remote OpenAI-compatible NIM or NVIDIA-hosted endpoint for agentic retrieval. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Supports the Super-49B default LLM path and the Omni VLM-capable answer path. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Is supported across in-process and batch ingest modes. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Configures the worker acknowledgement timeout for VectorDB writes; defaults to 300 seconds. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Provides built-in summarization with either the default prompt or a custom prompt. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Adds a shared one-request-per-row text-generation operator for graph workflows. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Configures text-generation operators with supported model, endpoint, generation, prompt, retry, timeout, and worker settings. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Adds a shared one-request-per-row text-generation abstraction for graph workflows. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Adds service-mode chunking for TXT and HTML inputs. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)

### Performance

- Runs OCR crop batches across page rows to improve throughput. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)

### Behavior and default changes

- Each write commits rows independently; index maintenance is serialized and coalesced, and expired index-readiness waits leave committed rows queryable. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Changing a Helm NIM image repository or tag on an existing release cannot patch the modelPuller source; delete the NIMCache and its PVC, then upgrade. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- The self-hosted default answer LLM is nvidia/llama-3.3-nemotron-super-49b-v1.5. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Helm defaults to Nemotron OCR v2, using the nvcr.io/nim/nvidia/nemotron-ocr-v2:2.0.1 image. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Helm maps rerank endpoint and model settings and auto-wires them when the rerankqa NIM is enabled, enabling reranking in split topology. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Helm's default VL embedding NIM image is upgraded to 2.3.0. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Helm's default VL reranking NIM image is upgraded to 2.3.0. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Nemotron 3 Nano Omni becomes the default image-captioning model for local and hosted paths; chart-classified PDF regions remain on layout and OCR. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)

### Tooling

- Development Docker Compose uses the combined object-detection image, with 2.0.0 tags unless NIM_*_TAG is overridden. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)

### Fixes and other changes

- Fixes local Hugging Face batch audio extraction hanging in interactive terminals when FFmpeg inherited the parent process's standard input. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Reports the package version instead of a stale package-version value, with Helm setting it from the running service image tag. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Agentic retrieval remains opt-in in service mode and requires a remote OpenAI-compatible LLM endpoint. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)
- Inherits from TextGenerationTask. (26.08.1) — [release notes](https://github.com/NVIDIA/NeMo-Retriever/releases/tag/26.08.1)