Select a database in the main menu to view report packages.
Report Summary
Loading…
Files
Generated Blocks
Query blocks and logic blocks generated from this report. Click "Generate Blocks" above to extract them from the report's summary and SQL scripts.
Table linkages
After report analysis and Generate Blocks, this section shows how tables are linked to each other and to query blocks, link blocks, logic blocks, and recipes.
Report Summary
Back to packageFiles in this package
AI Summary
No summary yet. Click "Generate summary" to analyze .txt files in this package and create an AI-generated summary. The summary will be stored as report-summary.md in the report package folder.
Generated Blocks
Query blocks and logic blocks generated from this report. Click "Generate Blocks" to extract them.
Files
Select a database in the main menu to view files.
Select the database used for queries. This is synced with the sidebar.
Connections
Saved connections can be used when importing from database. After import, the database is linked and shown as Live.
| Name | Server | Database | Actions |
|---|
Available databases
| Database | Tables | DB type | Live | Enabled | Actions |
|---|
Provider Status
Test connectivity to each LLM provider. Click Test to authenticate and fetch available models.
| Provider | API Key | Status | Models |
|---|
Default fallback
Used when an agent or tool has no override. Also used for the model dropdown below.
| Configuration | Provider | Model | Max tokens | Temperature |
|---|---|---|---|---|
| Default fallback |
LLM by agent
Optional overrides per query agent (Fast, Thinking, Recipe, Research). Empty = use default.
LLM by tool
Optional overrides per tool (intent, SQL generation, extract, etc.). Empty = use default. If all tools show the same provider, they may have been saved with that value; use "Reset tools to default" to clear overrides so they inherit the default above.
Saved to this browser. Per-agent and per-function LLM apply to queries and metadata actions.
Prompt Factory
Manage all LLM prompt templates. Each process step has its own prompt with six sections. Click a row to view and edit.
| Process | Category | Agent / Used By | Output Type | Placeholders | ||
|---|---|---|---|---|---|---|
| Loading prompt templates... | ||||||
Domain management
Define and describe business domains for the selected database. Each domain can have description line items that explain what the domain covers.
New Domain
Add line items that describe what this domain covers. Each line is a separate description entry.
Phrases that represent key concepts for this domain (e.g. "Prospect / lead tracking", "Commission tracking & splits").
Master data
Define Table classification, Table type, and Tags for the selected database. These values are stored in _master_metadata.yml and used by the Metadata Editor (tables and fields). Domains are managed in the separate Domain management section.
Domain-specific labels for tables (e.g. Tenant, Property). Used when editing table metadata.
Categories for how tables are used (e.g. master, fact, dimension). Helps organize and describe your schema.
Intent keywords for filtering and ranking tables (e.g. ar, cashflow, tenant_charges). Helps the AI pick the right tables for a question.
Use “Populate from tables” to fill the lists from current domain, classification, table type, and tags in _master.yml.
Audit log
Who did what, when. Filter by actor, action, or resource.
| When | Actor | Role | Action | Resource | Outcome | Duration | Details |
|---|
Usage summary
Input and output tokens by period. Cost is estimated using separate input/output prices per provider below.
Top token consumers (by call type)
Which tools and agents use the most tokens. Use this to find the most token-hungry tasks.
| Call type / Tool / Agent | Calls | Input | Output | Total | Cost (USD) |
|---|
Token usage stats by Tools
Minimum, maximum, and average input/output tokens per call for each pipeline tool (last 30 days).
| Tool | Calls | Min input | Max input | Avg input | Min output | Max output | Avg output |
|---|
Token usage stats by Agents
Minimum, maximum, and average input/output tokens per call by feature (e.g. query_generate, query_stream; last 30 days).
| Feature / Agent | Calls | Min input | Max input | Avg input | Min output | Max output | Avg output |
|---|
Usage log
Per-call token usage (latest first, 10 per page). Filter by range, user, feature, call type, or provider.
| When | User | Feature | Call type | Provider | Model | Input | Output | Total | Cost (USD) |
|---|
Token costs by provider
Set price per 1 million input and output tokens (USD). Cost per call = (input_tokens/1M × input price) + (output_tokens/1M × output price). Leave 0 for free (e.g. Ollama).
| Provider | Input (per 1M tokens) USD | Output (per 1M tokens) USD |
|---|
Clear usage history
Permanently delete all LLM usage log data (tokens, costs, call types). This cannot be undone. Provider cost settings above are kept.
Conversations with LLM call details
Chats that have logged LLM calls. Click a row to view the full conversation and every prompt/response/reasoning for that chat.
| Question(s) | LLM calls | Last call |
|---|
Conversation and LLM call details
LLM calls (prompt, context, response, reasoning)
Training runs with LLM call details
LLM-based training jobs (report packages, Learn, analyze tables, etc.). Click a row to view every prompt/response/reasoning for that run.
| Process type | Entity | LLM calls | Last call |
|---|
Training run and LLM call details
LLM calls (prompt, context, response, reasoning)
LLM training queue
FIFO queue for training jobs: Learn from live (Settings), Report package Analyze / Generate summary / Generate blocks (Reports). Chat and agents run with priority. Jobs persist across restarts.
| ID | Type | Status | Created | Started | Finished | Result / Error |
|---|
iDrive E2 (S3-compatible) storage
File uploads, downloads, exports, and temp files are stored in iDrive. Configure credentials below. Create a bucket in iDrive E2 first.
User management
Add, edit, or remove users. Only admins can access Settings and Metadata Editor.
Users
Common Reports
Reports use the database selected in the main menu (sidebar). Template reports with execution plans (Markdown) for the LLM to build each report. Import docs/common-reports.csv to seed the library.
| Domain | Report name | Report description | Actions |
|---|
Report
Execution plan
Detailed instructions (Markdown) for the chat agent to build this report. View rendered below; click Edit to change, or use Generate to create a draft from the report name and description.
No execution plan. Click Edit to add one or Generate to create a draft.
Table attributes
Foreign keys & relations
Columns
| Column name | Datatype | Description | PK / FK |
|---|