Documentation V1

August 14, 2026

Updated August 15, 2026.

Documentation center#

Added structured Overview, Agents, API Reference, and Changelog navigation.
Added full-text search, desktop and mobile reading layouts, raw Markdown pages, an AI-readable documentation index, canonical metadata, and sitemap coverage.
Added cross-links between available guides, troubleshooting pages, and API operations.

Added guides for the text-agent lifecycle: create and configure an Agent, test a frozen Draft, publish immutable Versions, manage Branches and deployment traffic, inspect Conversations and Activity, and administer Workspace roles and credentials.
Added dashboard coverage for prompts, models, Conversation settings, Knowledge, Tools, Procedures, Workflows, Guardrails, Analysis, Branch history, Deployments, monitoring, usage and quota, members, API keys, Secrets, and post-call webhooks.
Added a reviewed /api/v1 reference with method, path, authentication class, minimum role, API-key availability, request and response schemas, route-declared errors, and canonical examples.
Added API foundation guides for authentication, headers and content types, Workspace scope, errors, filtering and pagination, concurrency, post-call signatures, and breaking-change notices.
Added twelve production guides for Agent Testing: library and sharing, three test types, conversation prefill, Main Draft and Live runs, results, durable recovery, usage safety, and troubleshooting.
Expanded the reviewed API reference to 64 paths and 89 operations, including all Agent Test, folder, invocation, run, resubmit, bucketing, and conversation-prefill routes.

Operade Agents are documented as text-based.
Telegram is marked Limited because the HTTP receiver exists, webhook registration is not automated end to end, and the polling utility is a developer alternative.
Post-call webhook delivery covers successful Analysis results, signed raw-body verification, and bounded retries.
Troubleshooting guides separate symptoms, causes, verification, recovery, and escalation.

Added a dashboard guide for connecting an explicitly selected My Drive spreadsheet, configuring bounded Search rows or Add one row actions, testing real append side effects, and cloning a capability after an Agent version locks it.
Added troubleshooting and canonical error references for account, file, schema, quota, response-size, availability, and uncertain append outcomes.
Added reviewed dashboard-session API metadata and operator guidance for feature-flag rollout, silent credential preflight, path-only logs, QA/production isolation, and forward recovery after data exists.
The integration remains behind an operator-controlled feature flag, while provider testing stays limited to approved canary accounts. Production enablement, external OAuth consent-screen publication, and privacy text require separate owner and legal approval.

Begin with the Dashboard quickstart or API quickstart. Developers can continue to the API introduction, while operators can use Monitor troubleshooting and Support.