3. Your first chat message

~4 minute read. Two minutes to a streaming answer grounded in your own data.

Step 3: Open e.chat

Go to https://app.eworks.cloud/chat.

[Screenshot: e.chat with history panel on the left, model selector top right, message box at the bottom]

Orientation

• History panel (left) — every conversation in this workspace that you have access to, newest first, full-text searchable.
• Model selector (top right) — Claude, GPT, Gemini, and any self-hosted Ollama models your admin has registered. Auto lets e.brain route each message to the cheapest model that can handle it.
• Grounding chip — pick a knowledge collection to answer from. With none selected, the model answers from its own training data only.
• Message input (bottom) — Enter sends, Shift+Enter adds a line, / opens the command palette, and you can drag files straight in.

Send a message

1. Select Claude in the model selector.
2. Type What is AI? and press Enter.
3. Watch the response stream in token by token. Stop it any time with Esc.

That's it — you've made your first billed call. The cost appears under the message (typically a fraction of a cent) and lands in e.dash and the audit trail within seconds.

Ground it in your own content by picking a collection from the grounding chip and asking something only your documents would know. Answers then come back with inline citations you can click through to the source.

Export a conversation

⋯ → Export on any conversation:

• PDF — formatted transcript with timestamps, model names, and citations. Good for sharing and for evidence packs.
• JSON — full fidelity, including tool calls, token counts, and per-message cost. Good for pipelines.
• Markdown — for pasting into tickets and wikis.

curl -s https://api.eworks.cloud/v1/chats/$CHAT_ID/export?format=json \
  -H "Authorization: Bearer $ACCESS_TOKEN" -o chat.json

Session persistence

Conversations are saved to your workspace as you type — no "save" button, and closing the tab loses nothing. They stay until someone deletes them or your workspace retention policy expires them (default: keep forever). Everything is searchable from the history panel, and shared conversations are visible to the teammates you shared them with, subject to their role.

Troubleshooting

"The model selector won't load." It's fetching the model catalog from e.brain. Reload once. If it's still empty, no models are enabled for your workspace — an Admin fixes that in Settings → Models. A corporate proxy blocking api.eworks.cloud produces the same symptom.

"My message didn't send." Look for a red retry icon on the message. Common causes: the connection dropped mid-stream (click retry — nothing was charged for the incomplete part), the selected model was disabled while you were typing (pick another), or your input exceeded the model's context window (shorten it or start a new conversation).

"I hit a rate limit." Beta workspaces are capped per minute and per day. The error tells you which limit and when it resets. Switching to Auto routing usually clears short bursts, since it spreads work across providers. Admins can see current consumption in e.dash → Usage and request a raise from support.

Next: Create your first agent