Roomote

Roomote stores conversations and delegated tasks on its server. Hansard offers two explicit import paths for a deployment.

Source IDs

Source	Use it for
roomote-db	PostgreSQL conversation history, inference usage, and retained memory text.
roomote-cloud	Authorized visible transcripts through the Roomote API.

How the import works

The database reader asks PostgreSQL for a consistent, read-only copy of specific history tables and columns. It preserves primary messages, tool arguments and outputs, task attempts, and durable inference usage. Task usage belongs to the task even when the same row also names its parent session. Selected model settings remain metadata; only recorded inference models enter model accounting. Connection credentials and unrelated deployment settings are excluded from the archived copy.

The cloud reader enumerates accessible sessions, reads their summaries and linked task transcripts, and verifies that summaries remain stable during capture. It requests the transcript endpoint without a message limit to avoid the separate session endpoint’s thousand-message cap. This API returns sanitized visible text and image references, so it cannot supply the database reader’s structured tool arguments or inference usage. Both paths derive the same identities from the deployment origin, allowing richer database imports to replace earlier cloud captures without duplicating conversations.

Import and keep current

Conversation history

Set ROOMOTE_URL to the deployment’s HTTPS origin. Loopback HTTP is also supported for a local server. Supply ROOMOTE_DATABASE_URL and install psql for database imports; HANSARD_ROOMOTE_PSQL can select its executable.

hansard import roomote-db --since all

For cloud imports, supply an authorized bearer token in ROOMOTE_AUTH_TOKEN:

hansard import roomote-cloud --since all

Both commands accept --session <uuid>, --dry-run, and --json. Use the same ROOMOTE_URL for both paths. Repeat the command to capture subsequent turns; these sources remain outside the local watcher and source picker.

Memories and instructions

hansard import roomote-db --since all

The database import also snapshots deployment instructions, task instructions, and retained Fast/Brain memory outbox text. Task memories retain repository scope when an explicit repository URL is present. These observations are read-only and may differ from finalized Brain pages. Periodic memory backup does not query Roomote, and the cloud transcript path does not export memory.

Review the import in the app

1. When the import finishes, open Conversations and selectFilter & sort.
2. Set Source to Roomote and setTime to the window you imported. UseFolder to narrow the list when you use this harness in several projects.
3. Select a conversation to check its messages, tool calls, files, commits, model, and token usage.

To keep archiving new sessions from a local harness, openSettings → Imports. Automaticfollows every supported harness installed on this computer.Custom lets you turn individual harnesses on or off. Cloud sources and account exports never run from here; each needs its own import command. If a conversation you expect is missing after you reset the filters, run the import command on this page again with --explain-skips where the source supports it.

Check database tool arguments and edit diffs, parent-task links, recorded inference models, and memory scopes. Cloud tool previews remain context text.

What Hansard preserves

• Database messages, reasoning, tools, results, and inline attachments.
• Durable inference tokens, cache and reasoning counts, and reported costs.
• Delegated task links, resumed task attempts, and automation classification.
• Sanitized cloud transcripts and provider summaries returned by read endpoints.
• Raw history copies or API responses, plus read-only memory snapshots.

Known limitations

The supported schema is Roomote 1.13.1. Database imports include standalone tasks and sessions, including retained archived records. Legacy compatibility message arrays remain raw-only; Fast conversations without a session record are outside this import. Encrypted personalization, runtime secrets, compute files, and finalized Brain pages are not exported. Remote attachment bytes are not fetched, and upstream tool truncation cannot be reversed.

Cloud listings omit hidden and archived sessions. An authorized known session can be requested explicitly, subject to server access controls. Cloud reads have no atomic snapshot; a changed summary rejects that session for retry. Captures are limited to 2,000 sessions, 100,000 database rows or messages per cloud transcript, and 128 MiB. Select a session when a limit is reached. Authenticated hosted inference, delegation, and cloud acceptance remain pending.

Refresh an existing archive

Repeat the corresponding hansard import roomote-db --since all or hansard import roomote-cloud --since all command after a parser update. Subsequent cloud imports retain an existing database archive of the same identity. Provider originals are required; raw replay is not implemented.