Agnos Launches Auditable Local AI Document Cleaning

// 3h agoPRODUCT LAUNCH

Agnos Launches Auditable Local AI Document Cleaning

Agnos is a Windows desktop tool that removes document noise—headers, footers, page numbers, cookie banners, boilerplate, and duplicates—before files enter RAG systems, chatbots, or training datasets. It processes files offline, shows the rule behind every removal, and exports cleaned text, JSONL chunks, or training data.

// ANALYSIS

Agnos addresses an overlooked source of poor AI quality: messy source documents contaminating retrieval and training pipelines. Its audit trail is more compelling than fully automated cleaning, but the Windows-only footprint and lack of OCR limit its reach.

–Local, offline processing is valuable for sensitive legal, financial, healthcare, and internal company documents.
–Reviewable removal rules make cleanup safer than opaque preprocessing scripts or cloud APIs.
–Its study found page furniture was only 1.45% of text by character count, but repeated placement can still pollute chunk retrieval.
–The reported false positives show human review remains necessary for tables, phone numbers, labels, and requirement numbers.
–A $99 one-time license is attractive for individuals, though teams needing Linux, macOS, OCR, or server-side automation will need alternatives.

// TAGS

agnosragdata-toolslocal-firstdevtooltraining

DISCOVERED

3h ago

2026-10-31

PUBLISHED

8h ago

2026-10-31

RELEVANCE

8/ 10

AUTHOR

Michael Shadpour

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE23m ago

MCP TypeScript SDK adds built-in OpenTelemetry

The official MCP TypeScript SDK merged built-in OpenTelemetry tracing and duration metrics for clients and servers, targeting the next version bump. Trace context propagates through params._meta, while tool arguments and results stay out of spans by default.

RESEARCH42m ago

MAGE Bridges Multimodal Recommendation Gaps

MAGE reshapes text embeddings with complement-aware residuals, then uses an asymmetric dual-stream graph neural network to connect visually similar products with functionally complementary ones. The open-source implementation includes preprocessing, datasets, training, and evaluation code.

RESEARCH51m ago

IBM JevAct Adds Decision Gates to Agents

IBM SIL’s JevAct inserts a runtime verification layer between tool-using agents and execution, combining executable policy checks with probabilistic yes/no judgments and bounded revision. Tests across τ²-bench and BFCL improved reliability, especially on Telecom tasks.