emrg: memory reflection tool log — session id + round + truncation marker - #815
Closed
argszero wants to merge 1 commit into
Closed
emrg: memory reflection tool log — session id + round + truncation marker#815argszero wants to merge 1 commit into
argszero wants to merge 1 commit into
Conversation
…tion marker Host rant 2026-08-17T12:00:35: 'memory reflection tool: bash → ---' was unclear — no context (which background task), no session id (other lines in the same function carry id=%s round=%d), and the truncated '---' looked like meaningless output rather than the first 100 chars of a YAML frontmatter. - Prefix unified to 'memory reflection:' (matches sibling logs) - Added id=<session_id> round=<n> (the for-loop round, 1-based) - Result truncated to 100 chars with a '…' suffix when longer, so it is unambiguous that the output was cut. Example: memory reflection: id=s_xxx round=2 tool bash → ---… No test changes (log-only).
argszero
commented
Aug 17, 2026
argszero
left a comment
Owner
Author
There was a problem hiding this comment.
✅ LGTM — cycle (self-review): log-only change; prefix unified to memory reflection: with id/round, 100-char truncation gains a … marker. Sibling logs already carry id/round — consistent. No test impact.
Owner
Author
|
Superseding this PR: the same memory-reflection log fix (id/round/truncation marker) is now contained in #814 (feature/submit-rant-tool), which additionally adds the tool |
argszero
pushed a commit
that referenced
this pull request
Aug 17, 2026
… 2026-08-17T12:03:13)
Host: every agent tool log should output 'tool name + human-readable
purpose' so background calls (memory reflection / consolidation) are
understandable without context. Also submit_rant's project becomes
REQUIRED (host 12:00 demand folded into this rant).
- ToolDefinition gains purpose: human-friendly one-line purpose
(logs/UI); description stays for LLM routing.
- All 7 tools (bash/read/write/edit/glob/grep/submit_rant) get a purpose.
- Log sites now print 'name — purpose (args)':
* main loop: 'tool call: bash — 执行 shell 命令并返回输出 ({command: ...})'
* memory reflection: 'memory reflection: id= round= tool name — purpose → out…'
* consolidation: 'consolidation tool: name — purpose → out…'
(unknown tool → purpose 'unknown tool'; tool mocks in tests updated)
- submit_rant: parameters.required now ['project','message']; execute
rejects empty project with 'ask the user which project this rant targets'.
- system.j2 Rant Handling: tool requires project — ask user if unknown.
Also covers rant 2026-08-17T12:00:35 (memory reflection log id/round/
truncation marker) — same log line, superset, #815 closed.
+2 tests (project-required, all-tools-have-purpose), Agent.md 841→843.
Contributor
|
I tested PR #815 end-to-end (checked out
One non-blocking observation for a future slice: the |
argszero
added a commit
that referenced
this pull request
Aug 17, 2026
…hen-write) (#814) * emrg: submit_rant tool + rant in normal conversation (rant 2026-08-17T11:51:59) Host: a rant is not a special mode — it is a normal part of conversation. The user may complain/suggest in plain speech; the agent should detect rant intent, clarify/polish, get explicit consent, then write it. - New emrg/server/rants.py: append_rant() shared write logic (field order timestamp → project → status → progress → completed → message, tz-aware daemon-authoritative timestamp, sorted rewrite) — extracted from the daemon rant handler, single source of truth. - New emrg/tools/submit_rant_tool.py: SubmitRantTool registered in the daemon tool registry (available in every session); description mandates explicit user consent before calling; project optional (default = emrg). - daemon.py rant handler now delegates to append_rant (identical behavior). - system.j2: new 'Rant Handling' section — recognition (no /rant prefix needed), confirm → clarify → polish → show → call submit_rant; never call without explicit agreement; /rant <msg> and GUI panel are explicit. - TUI /rant <msg> / /rant @Proj <msg> no longer write directly: routes through the agent as a normal task with a '[Host wants to submit this rant…]' hint (agent confirms + calls submit_rant). GUI panel keeps the direct write (explicit form submit = confirmed). - +7 tests (append_rant sort/field-order/corrupt-skip, tool write/count, empty-message error, definition consent contract, system.j2 section, tool registered). Agent.md 834→841. * emrg: tool purpose field in logs + submit_rant project required (rant 2026-08-17T12:03:13) Host: every agent tool log should output 'tool name + human-readable purpose' so background calls (memory reflection / consolidation) are understandable without context. Also submit_rant's project becomes REQUIRED (host 12:00 demand folded into this rant). - ToolDefinition gains purpose: human-friendly one-line purpose (logs/UI); description stays for LLM routing. - All 7 tools (bash/read/write/edit/glob/grep/submit_rant) get a purpose. - Log sites now print 'name — purpose (args)': * main loop: 'tool call: bash — 执行 shell 命令并返回输出 ({command: ...})' * memory reflection: 'memory reflection: id= round= tool name — purpose → out…' * consolidation: 'consolidation tool: name — purpose → out…' (unknown tool → purpose 'unknown tool'; tool mocks in tests updated) - submit_rant: parameters.required now ['project','message']; execute rejects empty project with 'ask the user which project this rant targets'. - system.j2 Rant Handling: tool requires project — ask user if unknown. Also covers rant 2026-08-17T12:00:35 (memory reflection log id/round/ truncation marker) — same log line, superset, #815 closed. +2 tests (project-required, all-tools-have-purpose), Agent.md 841→843. * test: append_rant sort test uses host-local offset (UTC CI fix) The older-rant fixture hardcoded a +08:00 timestamp; on UTC CI hosts the daemon writes +00:00, so the lexicographic string sort puts the new entry first and the order assertion failed. Production is unaffected (all rants share the host offset), but the test must too — compute the older timestamp as now - 1h in the host's own tz. Verified under TZ=UTC and TZ=Asia/Shanghai. --------- Co-authored-by: EMRG Evolution <emrg@argszero.dev>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Host rant 2026-08-17T12:00:35:
memory reflection tool: bash → ---in emrgd.log is unclear — no context (which background task), no session id (sibling logs carryid=%s round=%d), and the truncated---looks like meaningless output rather than the first 100 chars of a YAML frontmatter.Change (log-only, 6+/1-):
memory reflection:(matches sibling logs in_maybe_reflect_memory)id=<session_id> round=<n>(1-based tool-loop round)…suffix when longer — unambiguous that the output was cutNew form:
memory reflection: id=s_xxx round=2 tool bash → ---…No test changes (Agent.md count unchanged).