Topic
AI and agents
Company data as context for models, retrieval and autonomous agents.
- Company data as context for AI agents and copilotsDesign the tools, the result format and the tests for an agent that answers questions about companies from dated, sourced rows instead of scraped pages.
- Retrieval-augmented generation over structured company signalsMost questions about companies ask for a count, a date or a list, which a database answers exactly. This guide shows what to embed, what to filter and what to leave to SQL.
- Training and evaluating models on dated company signalsA model on company signals is only as honest as its feature table. This guide builds one as of each date, splits by time and names the properties of the record that keep it safe.
- Company signals as context for Amazon Bedrock applicationsCompany signals are dated rows with sources. This guide compares an Amazon Bedrock knowledge base with a tool call over your own tables, and shows how to keep every answer dated and cited.
- Company signals as grounding data for Vertex AIWhich of Google's grounding options fits company signals, how to expose BigQuery tables to a Gemini model with typed queries, and how to return dates and sources so an answer can be checked.
- Feature engineering on company signals in DatabricksWhich timestamp to key a feature table on, how to join labels as of a date, and the leakage traps left in daily and weekly company data, with worked SQL and Python.
- Grounding Azure AI applications on company signalsTwo routes to answers that carry a date and a source: a search index for announcements and a SQL tool for counts and trends, with the tables, the tool definition and the citation rules.
- Serving company data to AI agents over Model Context ProtocolA team wraps the Fokals REST API as MCP tools in a thin server of its own. This guide covers tool design, paging, keys, rate limits and untrusted text against the current specification.
- Using company signals with Snowflake Cortex AI functionsWhich Cortex functions exist today, worked SQL to summarise a watchlist's announcements and to search them, and the governance, cost and licence questions to settle first.