API reference / Tasks and projects
Write down what a task taught
Asks the teammate of a task to write down what the task taught it, so that it needs a person less next time. The teammate reads where a person stepped in: answers to its questions, calls that a person turned down, comments, and work that a person sent back. It writes 2 lessons at most: a skill of the workspace, or one line for its own notes. Each lesson waits as an approval with lesson set. Answers 202 with the run; poll statusUrl. Troopy also does this by itself after a task is done, unless learnFromWork is off.
/api/v1/tasks/{key}/lessonsOther ways to call it
troopy tasks learntroopy_learn_from_taskIdempotency-Key header. A retry with the same key gets the first answer back, and the operation runs one time.Path parameters
key*stringA task's key (VIR-12), its number (12) or its id
At least 1 characters · At most 40 characters
Response
202 · application/jsondata*object21 properties
id*stringA 24-character id
agent*stringkind*enum<string>compile_profile: the lead writing the short form of the long memory files; summarize: a teammate keeping the summary of a long conversation; translate: the Support agent translating the messages of a support chat, for Live translate; dream: the lead linking the notes of the brain, and inferring a few facts from them; learn_facts: the lead reading the facts of docs, web pages and files into Memories; watch_sessions: a teammate reads the sessions of visitors, and names the problems that they show; draft_brand: the lead drafting the empty lines of the brand from its website; classify_feedback: a teammate suggests the type and the themes of what customers said; fill_column: a batch of a list's AI column; triage: the lead's quick read of a troop chat message naming no teammate; distill: the lead reading what people said, for facts to keep: what they said to teammates, and their own talk with learnFromTalk; write_emails: a batch of a campaign's emails, written for each person; lessons: a teammate writing down what a finished task taught it (learnFromWork); write_rebuttal: a teammate writing the answer to a Stripe dispute, which Troopy checks against the records; read_photo: the lead reading the notes of a photo of a board onto a canvas
workdraft_roledraft_memorydraft_skilldraft_selffill_columntriagedistillwrite_emailslessonswrite_rebuttalclassify_feedbackdraft_brandread_photowatch_sessionslearn_factsand 4 more
trigger*object5 properties
type*enum<string>What started it: a message, a routine, a teammate handing work over, drafting a role, setting up memory after onboarding, a task, a message in the troop chat, a list's work column, the emails of a campaign, incidents to look at, a dispute to answer, what customers said to sort, a look at the sessions of visitors, documents whose facts wait for a read, a photo of a board to read onto a canvas, or a support chat with Live translate on
chatroutinescreenhandoffhiresetuptasktroopcolumncampaignincidentdisputecustomerssessionsbraincanvasand 1 more
threadId*string | nullroutineId*string | nullfromRunId*string | nullThe run that handed this work over
troopThreadId*string | nullA message in the troop chat: the thread it answers in, named by the thread's first message
taskId*string | nullThe task it worked on
task*stringstatus*enum<string>queued or running: poll statusUrl every few seconds
queuedrunningdonefailed
roleVersion*integersummary*stringOne line on what it did
error*string | nulldoing*string | nullWhile it works: what it is doing, in its own few words, like "Checking Acme's payment terms"; null before its first step says
toolCalls*integerskills*array<string>The skills it used, by name: asked for with /name at the start of a message, then opened
requestedBy*objectWho wrote it
3 properties
type*enum<string>personagentapi
id*stringname*stringcreatedAt*stringISO 8601 date and time
finishedAt*string | nullstatusUrl*stringreply*string | nullIts answer, in markdown
cites*array<object>A teammate's answer: what in the brain it relied on, at most 5, each one it was shown, as it is now (what was deleted since is left out)
5 properties
type*enum<string>docsource
id*stringA doc's slug, else the id
title*stringkind*stringA doc's or source's kind: a note, a web page, an interview, a research report...
wrong*object | nullA note someone said is wrong, and who: out of search until restored, or until they take it back
3 properties
by*objectWho wrote it
3 properties
type*enum<string>personagentapi
id*stringname*stringat*stringISO 8601 date and time
userId*stringA 24-character id
draft*object | nullFor draft_role runs: the drafted role, to hire with POST /agents
rounds*array<object>5 properties
startedAt*stringISO 8601 date and time
endedAt*string | nullaction*enum<string> | nulltoolsfinish
doing*string | nullWhat the round did, in a few words of the teammate. Null for an older round
calls*array<object>7 properties
tool*stringlabel*stringinput*anywhy*string | nullWhy the teammate made the call, in its own words. Null when the run did not record it
status*enum<string>waiting: stored for a person's approval. refused: not allowed, or invalid input. running: it runs now
donefailedwaitingrefusedrunning
result*stringWhat the teammate read back, cut to a size. Empty while the call runs
approvalId*string | null