Fieldy

Use this for recent activity, date ranges, complete lists and counts: "what did I record today?", "recap my week", "every meeting with [email protected]", "how many client visits did I have in September?". Lists conversations newest first, with optional attendee name or email, calendar title, recorded location or venue, recording source and date filters, read from stored metadata without the search index. Follow nextCursor until hasMore is false; count is only this page. An attendee filter matches calendar invitations, so report totals as conversations with that person invited, attendance unverified; claim an attendance total only after checking every counted conversation's speakers or transcript, since a sample cannot establish it. Open a conversation with fieldy_get_conversation for its contents. Do not use to find a topic or phrase; that is fieldy_search_conversations. coverage.account.accessibleFrom is the account's history boundary; earliestConversationAt and latestConversationAt describe stored history, not accessible history. Requests before the boundary are inaccessible, not evidence of no meetings, and no other tool extends this access.

fieldy_browse_conversations

Use this when you have a conversation id and need what happened in it: "summarize that meeting", "what were the action items?", "what did we decide?", "draft a follow-up email from it". Returns the full record: title, short and long summary, content, time range, keywords, quotes, recorded location, speakers, recording source, a private Fieldy link, linked calendar events (meeting titles, invitee names and emails, RSVP status, organizer, invited attendee counts, scheduled venues), template metadata, and `transcript`, the complete transcript, never truncated. `summary` and `content` can be null when no summary was generated; the transcript is the ground truth either way. `transcript` is null only when the token lacks `transcripts:read`. For one part of a very long conversation, fieldy_list_transcripts with a narrow window is cheaper. Do not use without an id: find it first with fieldy_search_conversations (topic, person, phrase) or fieldy_browse_conversations (dates, every meeting with someone, counts). Errors: - "Conversation <id> not found": wrong id, or the conversation was deleted - "insufficient_scope": token missing `conversations:read`

fieldy_get_conversation

Use this when the user wants exact words from a known time window: "what exactly did she say about the deadline?", "quote what I said at the start of the 10am meeting", "what was said in the last five minutes of the call?". Returns verbatim transcript lines in timestamp order, each with `speaker`, `text`, `timestamp`, and audio offsets. Also use it to verify a quote, attribute who said what, or read one slice of a long conversation. Do not use for: - One conversation's full contents: fieldy_get_conversation already embeds its complete transcript - Finding a conversation by topic: fieldy_search_conversations first Slicing recipes (combine `startTime`, `endTime`, `order`, `limit`): - Start of a window: `order: "asc"` (the default) with a small `limit` returns the earliest lines - End of a window: `order: "desc"` with a small `limit` returns the latest lines (re-sort ascending before showing them to the user) - Specific moment: set `startTime` and `endTime` to a tight interval around it (e.g. a few minutes) `endTime` defaults to "now" if omitted. Window may not exceed 7 days — narrower is much cheaper. Pagination: pass `nextCursor` as `cursor` for the next page; stop when `hasMore` is false. The cursor continues in the direction set by `order`, so keep `order` identical across pages. Errors: - "Invalid date format" — `startTime` / `endTime` could not be parsed - "Range exceeds N days" — narrow the window - "insufficient_scope" — token missing `transcripts:read`

fieldy_list_transcripts

Kept for older clients: lists full conversation records that started in [startTime, endTime), ISO dates, at most 30 days and 50 per page. Do not use for new requests; fieldy_browse_conversations with filters.from and filters.to answers "what meetings did I have last week?". To page, pass nextCursor as cursor until hasMore is false. Restricted content stays redacted.

fieldy_list_conversations_in_time_range

Kept for older clients: lists full conversation records newest first (default 20, max 50 per page). Do not use for new requests; fieldy_browse_conversations answers "what did I record today?", lists and counts with filters. To page, pass nextCursor as cursor until hasMore is false. Restricted content stays redacted.

fieldy_list_recent_conversations

Use this when the user asks what was said, discussed, decided or promised about a topic, person, company or phrase: "what did the client say about pricing?", "find the meeting where we talked about the Vilnius pilot", "what did Tom ask me to send?". Returns a ranked shortlist with matching transcript passages; open a result with fieldy_get_conversation for the full summary and transcript. Add date filters only when the user gives dates. The shortlist is not every match or a count; use fieldy_browse_conversations for "every meeting with Tom" or "how many". An empty result is not proof of absence, because older content may be missing from the search index: browse by date or attendee before concluding. coverage.account.accessibleFrom is the account's history boundary; earliestConversationAt and latestConversationAt describe stored history, not accessible history, and a request before the boundary is inaccessible, not evidence that no meetings exist. Transcript passages require transcripts:read.

fieldy_search_conversations

Use this when the user wants a link that someone else can open: "share the notes from my meeting with Tom", "give me a link to send to the client". Creates or re-enables the conversation's public share link and returns its URL. Anyone with the link can view the shared fields. Each conversation has one link; calling again updates it. It only creates the link and never sends it. Do not use to link the user to their own conversation; every record already has a private url.

fieldy_share_conversation