Roadmap
whodar answers "who do I talk to about X" from the data it has indexed. The next versions widen what it can see. Because every source is one connector against a single interface, each item below is additive and changes nothing downstream.
More sources
GitHub, Jira, Confluence, PagerDuty, and git history are shipped. Still to come, roughly in order of how many orgs they unlock:
- Microsoft 365: Teams messages and Outlook mail through one Graph API connector, so Microsoft-centric orgs get first-class coverage.
- Google Workspace: Docs, Drive, and Groups activity.
- Opsgenie: on-call schedules and service owners.
- Notion, Linear, and GitLab as the connector surface grows.
Each maps its data to people, teams, topics, and channels, and joins other sources by email or an alias file, so one person stays one entry across them.
Engine
- Binary vector store: embeddings are kept as JSON today, which is heavy for large organizations. A compact on-disk format keeps the index small.
- Incremental indexing: update only what changed instead of rebuilding.
Experience
- Result deep links: open a channel or a profile directly from an answer.
Hosted tier
Everything above assumes the self-hosted binary, which stays the default and the whole point: your data never leaves your walls. Alongside it, a separate hosted tier would offer a fully managed whodar you add to Slack in one click, for teams that want zero setup and are fine with a service running it for them.
This is a distribution model, not a feature of the binary. A one-click, directory-listed Slack app needs a public server to run the OAuth install, receive Events API callbacks, and hold each workspace's token and index. Socket Mode, the no-public-URL path the self-hosted bot uses, cannot be listed in the Slack Marketplace, so the hosted tier means a server in the loop by design.
That server changes the privacy posture, so the tier is explicitly opt-in and kept apart from the self-hosted promise. The self-hosted tool does not change: no account, no vendor, nothing sent out for review. The hosted tier is a choice for a different buyer, and the two never blur into one story.
Building it is real service work, not a small toggle: a public install and OAuth flow, per-workspace token and index storage, an Events API endpoint, Slack app review, billing, and on-call. It also needs a clear data-handling posture, since a hosted whodar processes the people graph on infrastructure the customer does not own. Scoped as its own track so the self-hosted core stays small and private.
Personal digest
A second way to ask: "what did I miss that matters to me." Instead of a topic query returning people, the digest scans the channels the running user can see, scores each message for personal relevance, and returns a ranked, actionable roll-up on demand, on a schedule, or live. It reuses the Slack client, the messaging client, the LLM layer, and the policy egress rules, and adds only a per-message relevance path beside the existing pipeline. See DIGEST.md for the full plan.