AI | Zparse Documentation

📄️Embeddings

Generate vector embeddings for each record in a JSONL file using a configured AI provider.

Vectorize a query attribute on a JSON record so it can be searched against a vector store.

Submit a JSONL batch job to an AI provider and stream results back as a file.

One-shot chat completion against a configured AI provider (OpenAI, Mistral, Anthropic, Gemini, Cohere, Moonshot, Scaleway AI).

Split JSONL records into smaller chunks using an LLM-driven chunking method.

Add LLM-extracted fields (sentiment, entities, classifications, …) to every record of a JSONL file.

Keep or drop each record of a JSONL file based on a natural-language acceptance criterion evaluated by an LLM.

Deterministically split text into fixed-size character chunks with optional overlap.

Extract a knowledge graph (entities + relations) from documents or JSON records using an LLM.

Expose connected Virtual Tools as a Model Context Protocol (MCP) endpoint that external MCP clients can call.

Split markdown text along heading boundaries into size-bounded chunks.

Submit a JSONL batch job to Mistral and stream results back as a file.

Send message to mistral chat

Generate embeddings

Generate embeddings search query

Run Mistral OCR on a document URL or file and emit page text plus optional structured annotations.

Generate embeddings using OpenAI's embeddings API.

Generate a single search vector via OpenAI's embeddings API.

Embed vector into Qdrant vector database

Generate embeddings search query

Deterministically split text by trying separators from coarsest to finest until chunks fit a size budget.

Re-score and filter a list of candidate documents against a query using an LLM/reranker provider.

Run a tool-calling LLM agent that picks among connected Virtual Tools to satisfy a prompt.

Declare a tool callable by a Virtual Agent or exposed via an MCP Endpoint, with typed argument outputs.