Talking to Agents
The Conversations page is where you interact directly with AEGIS AI agents. You can ask questions, request calculations, prepare filings, and get compliance advice through a chat interface.
Screenshot placeholder — to be added in a future update.
Accessing Conversations
Click Conversations in the sidebar to open the agent chat interface.
For filing preparation, the recommended workflow starts from the Compliance Dashboard rather than the Conversations page. The workspace-based approach provides a more structured, step-by-step experience. Use Conversations for ad-hoc questions, exploratory analysis, or when you prefer a free-form interaction.
Available Agents
AEGIS provides five agents you can converse with:
| Agent | Color | Best For |
|---|---|---|
| General Assistant | Purple | General questions about operations, data, or anything not specific to one domain |
| Rule 37 Filing Agent | Blue | Spacing exception filings under 16 TAC 3.37 |
| Rule 32 Flaring Agent | Amber | Flaring exception filings under 16 TAC 3.32 |
| Flaring Monitor | Orange | Volume tracking and emissions monitoring questions |
| Compliance Monitor | Green | Deadline tracking, risk scoring, and compliance status questions |
Starting a New Conversation
- Click the + New Conversation button at the top of the conversation sidebar.
- An agent picker panel appears showing all five agents with their descriptions.
- Click the agent you want to talk to.
- A new conversation is created and the chat area opens with a text input at the bottom.
You can also start a conversation from the empty state — when no conversation is selected, agent cards are shown in the main area. Click any card to begin.
Sending Messages
- Type your message in the input field at the bottom of the chat area.
- Press Enter or click Send to submit.
- The agent will process your message and stream its response in real time.
Voice Input
A microphone button next to the text input allows you to dictate your message using speech recognition. Click it to start speaking, and your words will be transcribed into the text field.
Understanding Agent Responses
As the agent works on your request, you will see several types of indicators:
Streaming Status
A bouncing dot animation appears while the agent is thinking, along with status text describing what it is doing:
- “Initializing agent…”
- “Loading memory context…”
- “Thinking…”
- “Selecting skills…”
- “Executing tools…”
- “Synthesizing response…”
- “Finalizing…”
Tool Execution Indicators
When the agent uses a tool (such as a spacing calculator or knowledge graph query), a blue status line appears showing what it is doing:
- “Calculating spacing distances…”
- “Querying knowledge graph for offset wells…”
- “Assembling Form W-1 filing package…”
- “Drafting good cause narrative…”
- “Calculating flaring volumes…”
- “Estimating CO2e emissions…”
HITL Banner
If the agent prepares a filing that requires review, a banner appears in the conversation:
“Filing package ready for review”
Click the Review button on the banner to go directly to the filing queue.
Agent Handoff
Sometimes an agent will suggest that a different agent is better suited for your request. When this happens, a handoff suggestion appears with a button to continue the conversation with the recommended agent.
Conversation Sidebar
The left panel shows all your conversations, grouped by agent type:
- Conversations are organized under their respective agent names with color-coded dots
- Click any conversation to reopen it and view its history
- New conversations that have not yet received a message are highlighted in blue
- Conversations awaiting HITL review show an “Awaiting review” badge
Collapsing Groups
Click the agent group header to collapse or expand conversations under that agent. This helps manage a long conversation list.
Conversation History
When you select an existing conversation, its messages are loaded from the server. You will see the full history of user messages, agent responses, tool status indicators, and any HITL banners.
Conversation Titles
The title of a conversation is automatically set from the first message you send (up to 50 characters). This helps you identify conversations later.
Tips for Effective Agent Conversations
- Be specific about the entity. Include the well name, API number, or lease name so the agent can look up the right data.
- State what you need clearly. For example: “File a Rule 37 spacing exception for Mitchell Ranch 1H” is better than “I need help with a filing.”
- Ask follow-up questions. If the agent’s response is not clear or you need more detail, just ask.
- One task per conversation. Start a new conversation for each distinct filing or analysis to keep context clean.
Example Prompts
Here are some effective prompts for each agent:
Rule 37 Agent:
“File a Rule 37 spacing exception for Mitchell Ranch 1H (API 42-329-40001-0000)”
Rule 32 Agent:
“File a Rule 32 flaring exception for the Mitchell Ranch lease”
Compliance Monitor:
“Show me the compliance status for our Permian Basin wells”
Flaring Monitor:
“What is the current flaring volume for Howard Unit A-1H and how close are we to the authorization limit?”