Knowledge Base | Product Updates | Mithrilis

Overview#

The Knowledge Base is Mithrilis's central data layer. It's the place where your team connects databases, manages business definitions, explores schemas, and queries data across systems — all without leaving the platform.

In logistics, data lives everywhere: carrier APIs, TMS platforms, warehouse systems, ERPs, spreadsheets. The Knowledge Base brings these data sources together into one browsable, queryable interface. It also feeds into Mithrilis's AI layer, so when you ask a question in the AI Assistant, it has the context to give you an accurate, grounded answer.

Data Source Management#

Connect external databases directly to Mithrilis. The system auto-detects existing database credentials from your configured integrations, so you don't need to re-enter connection details.

Schema Introspection#

Once a data source is connected, Mithrilis automatically discovers its structure:

Tables and views — Every table and view in the database is cataloged and browsable
Columns and types — Column names, data types, nullability, and default values are displayed
Relationships — Foreign key relationships between tables are detected and visualized
Caching layer — Schema metadata is cached for performance, so browsing is fast even for large databases

Multi-Database Support#

The Knowledge Base is architected to support multiple database types. Currently tested and working with PostgreSQL, with the infrastructure in place for additional databases as demand grows.

Automatic credential detection

If a database is already configured in your integration credentials, the Knowledge Base recognizes it automatically and offers it as a connectable data source. If not, it routes you to the credentials section to set one up.

Schema explorer showing table structure, column types, and foreign key relationships — Schema explorer showing table structure, column types, and relationships

Definitions Library#

A searchable library of business definitions scoped to each Knowledge Base. Definitions answer the question: "What does this term mean in our organization?"

Why Definitions Matter#

In logistics, the same term can mean different things to different teams. "On-time delivery" might be measured from pickup to final delivery by one team, and from warehouse departure to customer receipt by another. Definitions create a shared vocabulary that both humans and AI agents work from.

How to Use Definitions#

Create definitions for key business terms — "on-time delivery," "exception," "SLA breach," "proof of delivery," etc.
Scope definitions to a specific Knowledge Base, so different business units can maintain their own vocabulary
Search and browse definitions from the library interface
Definitions are automatically fed into the AI layer, so the AI Assistant uses your organization's terminology when answering questions

Start with your most ambiguous terms

When setting up the Definitions Library, start with the terms that cause the most confusion across teams. These are the terms where having a single, authoritative definition delivers the most value.

SQL Query Playground#

An interactive query interface for running SQL against your connected data sources. The playground went through two iterations to get the UX right, and the result is a tool your team can use daily to investigate data without needing a separate database client.

Features#

Write and execute SQL — Full SQL support against any connected data source
Syntax highlighting — Color-coded SQL keywords, table names, and strings for readability
Result table — Query results displayed in a sortable, scrollable table with column types
Query feedback — Rate query results (thumbs up/down) to improve future AI-generated suggestions
Saved queries — Bookmark frequently used queries for quick access. Saved queries are visible to your team.
Execution stats — See query execution time and row count for every run

SQL Query Playground with syntax-highlighted editor and sortable result table — SQL Query Playground with syntax highlighting and result table

Query Safety#

The playground runs queries in a read-only context by default. This prevents accidental data modifications when exploring. If you need write access for specific administrative tasks, it can be enabled per data source.

Large result sets

Queries returning more than 10,000 rows are automatically paginated. For very large exports, use the download button to get the full result set as a CSV file.

Documents#

Upload and manage documents that feed into the knowledge layer. Documents are processed and made available for AI-powered retrieval across the platform.

Supported Formats#

Upload PDFs, Word documents, plain text, and markdown files. Each document is:

Parsed and chunked for efficient retrieval
Embedded using vector embeddings for semantic search
Indexed so the AI Assistant can find relevant passages when answering questions

Use Cases#

Upload carrier rate sheets so the AI can answer questions about pricing
Store SOPs and operational procedures for team reference
Add contracts and service agreements for compliance lookups

Knowledge Base Nodes#

KB nodes (v1) plug directly into the Workflow Builder, letting automations read from and act on Knowledge Base data as part of any workflow execution.

How It Works#

Add a KB node to your workflow canvas, configure which data source and query to run, and the results become available as variables in downstream nodes. This means your workflows can make data-driven decisions based on live database content.

Example: A workflow that receives a carrier webhook can use a KB node to look up the carrier's SLA terms from your database, then use a conditional node to determine whether the event constitutes an SLA breach.

KB nodes and the AI layer

KB nodes are the building block that connects workflows to your data. Combined with AI nodes, you can build workflows that query your data, pass the results to an LLM for analysis, and take action based on the AI's output — all in a single automated pipeline.

Getting Started#

Connect a data source

Navigate to Knowledge Base from the sidebar, then click Add Data Source. If your database credentials are already configured in Mithrilis, the system will detect them automatically. Otherwise, enter the connection details manually.

Explore the schema

Once connected, click on the data source to browse its tables. The schema explorer shows column names, types, relationships, and row counts. Use this to familiarize yourself with the data available for querying.

Create definitions

Go to the Definitions tab and start adding business terms. Each definition should include the term, a plain-language explanation, and optionally a SQL expression or formula that defines it precisely.

Run your first query

Open the Query Playground, select a data source, and write a SQL query. Click Run to see results. Save useful queries for later access by your team.