mirror of
https://gh-proxy.org/https://github.com/klinsmaya/multica.git
synced 2026-11-08 17:36:45 +08:00
Bohan Jiangandmultica-agent
d54300abaa
Status leaves the compared set, so the snapshot is now just the two body fields. The comparison answers one question — must the agent re-read the issue BODY? — and status never helped answer it: its current value ships on every claim, so the agent learns it without a read. Worse, comparing it worked against the feature. The workflow has the agent set in_progress when it starts and in_review when it delivers, so the agent's OWN bookkeeping came back the next turn as "changed: status" and demanded a body re-read that nothing in the body justified. Replaying this issue's 21 follow-ups, that alone cut "unchanged" from 17 to 4. The earlier rationale for keeping it — a free signal that somebody intervened — was wrong on both counts: it is not free, and the agent can see an intervention by comparing the shipped status against the value it last set itself. issueSnapshotVersion stays 1, and this time rows in the old shape DO exist. That is still correct: removing a field is backward compatible, because every field the new code reads is present in an old row and still means the same thing, and the stale "status" key is ignored. The comment now states the actual bump rule rather than "the set changed" — bump when an old row can no longer answer the current question, which means adding a field or changing how a value is derived, not removing one. A test pins the legacy shape comparing correctly in both the unchanged and the changed direction, so that claim is checked rather than asserted. No migration: the column is JSONB and the schema is unchanged. TestClaimTaskByRuntime_AssigneeAndPriorityAreOutOfScope becomes NonBodyFieldsAreOutOfScope and now moves status alongside assignee and priority, asserting the current status still ships with its NEW value — without that, the fixture's UPDATE could silently not apply and the empty changed-list would prove nothing. Co-authored-by: multica-agent <[email protected]>
82 lines
4.2 KiB
Go
82 lines
4.2 KiB
Go
package execenv
|
|
|
|
import (
|
|
"fmt"
|
|
"strings"
|
|
)
|
|
|
|
// BuildIssueStateHint returns the workflow-step-1 pointer for a comment-
|
|
// triggered run: either the unconditional "read the issue" instruction, or —
|
|
// when the server actually compared the issue against this agent's previous run
|
|
// on it — what that comparison found.
|
|
//
|
|
// The problem it solves is the mirror of the comment delta's (MUL-6984, and
|
|
// MUL-7344 for the `--since` form). A follow-up turn that resumes its provider
|
|
// session already holds the issue body it read minutes ago, yet the per-turn
|
|
// message told it to run `multica issue get` again before doing anything. On an
|
|
// issue that did not move, that read returns bytes the session already has.
|
|
//
|
|
// Three renderings, and which one applies is decided by the CALLER, not here:
|
|
//
|
|
// - resumed && known && nothing changed → the comparison is reported as
|
|
// step 1's answer, and the read becomes conditional on the agent's own
|
|
// memory being insufficient. This is the only rendering that may replace
|
|
// the read, and it is allowed for exactly the reason the empty comment
|
|
// delta is: an affirmative server report about the whole record.
|
|
// - resumed && known && something changed → the changed field NAMES plus the
|
|
// read, so the agent knows what to look for rather than diffing blind.
|
|
// - anything else → the pre-existing instruction, byte for byte. A cold
|
|
// start, a dropped resume, an unknown delta and an old server all land
|
|
// here, and none of them may be softened: the agent has no prior context
|
|
// to continue from, or nothing looked.
|
|
//
|
|
// The compared set is named in full in the unchanged rendering rather than
|
|
// summarised as "the issue", because it is NOT the whole issue: labels, parent,
|
|
// due date, stage, project and metadata are not compared, and an agent told
|
|
// "unchanged" must be able to see that those were out of scope.
|
|
//
|
|
// Status and assignee ride along in both known renderings. They are what
|
|
// workflow step 3 needs ("is it already in progress", "is it mine"), and a run
|
|
// that is not reading the issue would otherwise have to read it for them.
|
|
//
|
|
// Like every other per-turn helper, this must never reach the runtime brief:
|
|
// its values are per-run, and the brief stays byte-identical across runs of a
|
|
// resumed session (MUL-5377).
|
|
func BuildIssueStateHint(issueID, issueStatus, assigneeType, assigneeID string, changedFields []string, deltaKnown, resumed bool) string {
|
|
readCommand := fmt.Sprintf("`multica issue get %s --output json`", issueID)
|
|
// The cold instruction is a byte-for-byte literal, not a format of the
|
|
// other two: it is what every non-resumed run has always been handed, and
|
|
// the tests that pin the cold path pin these exact bytes.
|
|
cold := fmt.Sprintf("Start by running %s to understand your task, then decide how to proceed.\n\n", readCommand)
|
|
// issueStatus empty means the server sent no issue state at all, which can
|
|
// only happen on a server predating these fields. Treat it as unknown even
|
|
// if deltaKnown somehow arrived set: reporting a status we do not have
|
|
// would be worse than an extra read.
|
|
if !resumed || !deltaKnown || issueStatus == "" {
|
|
return cold
|
|
}
|
|
state := fmt.Sprintf("status: %s; assignee: %s", issueStatus, issueAssigneeLabel(assigneeType, assigneeID))
|
|
if len(changedFields) == 0 {
|
|
return fmt.Sprintf(
|
|
"The issue is unchanged since your last run — the server compared title and description (%s). "+
|
|
"That answers workflow step 1: continue from your resumed context, and re-read with %s only if resumed memory is not enough.\n\n",
|
|
state, readCommand,
|
|
)
|
|
}
|
|
return fmt.Sprintf(
|
|
"Since your last run the issue changed: %s (%s). Read it: %s.\n\n",
|
|
strings.Join(changedFields, ", "), state, readCommand,
|
|
)
|
|
}
|
|
|
|
// issueAssigneeLabel renders the polymorphic assignee as one token pair. Both
|
|
// halves are required to mean anything — an assignee_type with no id, or an id
|
|
// with no type, is not an assignment — so a half-populated pair reads as
|
|
// unassigned rather than as a partial claim about ownership.
|
|
func issueAssigneeLabel(assigneeType, assigneeID string) string {
|
|
if assigneeType == "" || assigneeID == "" {
|
|
return "unassigned"
|
|
}
|
|
return assigneeType + " " + assigneeID
|
|
}
|