mirror of
https://github.com/THU-MAIC/OpenMAIC.git
synced 2026-10-17 05:29:46 +08:00
@openmaic/generation@0.3.7
17
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 LING DUANandwyuc
|
9a28ca39d5 |
feat(chat): add single PPT element references to Pi (#1224)
* feat(chat): add PPT element references to Pi * fix(chat): include chart values in reference routing * fix(chat): harden slide element reference evidence * fix(playback): clear element reference on scene navigation * fix(pi): preserve element grounding across child retries --------- Co-authored-by: wyuc <wang-yc24@mails.tsinghua.edu.cn> |
||
|
LING DUAN
|
146e164141 |
feat(whiteboard): add destructive runtime operations (#1173)
* feat(whiteboard): add destructive runtime operations * fix(whiteboard): preserve additive exact replay results * fix(whiteboard): support legacy code line ids |
||
|
LING DUANandwyuc
|
a0dccbe78d |
feat(pi): add Native Child additive whiteboard tools (#1156)
* feat(pi): add Native Child additive whiteboard tools * chore(renderer): bump package version to 0.1.3 * fix(pi): reject invisible whiteboard mutations * fix(pi): close native whiteboard review blockers --------- Co-authored-by: wyuc <wang-yc24@mails.tsinghua.edu.cn> |
||
|
LING DUAN
|
2cbe4ae293 |
feat(pi): add Native Child whiteboard tools (#1152)
* feat(pi): add native whiteboard tools * style(pi): format native whiteboard tests * fix(pi): clarify whiteboard CAS handoff * fix(whiteboard): reconcile runtime projection on open |
||
|
LING DUANandwyuc
|
59adbef03b |
Simplify Pi agent tool ownership and completion (#1148)
Co-authored-by: wyuc <wang-yc24@mails.tsinghua.edu.cn> |
||
|
LING DUANandwyuc
|
0d308ed78d |
feat(pi): add Native Child Web Search (#1133)
* feat(pi): add native child web search * fix(pi): align native web search with provider system * fix(pi): harden native web search configuration and cancellation * fix(pi): guide Native Child web search selection --------- Co-authored-by: wyuc <wang-yc24@mails.tsinghua.edu.cn> |
||
|
LING DUANandwyuc
|
cdf949a017 |
feat(pi): add native child runtime with scoped Spotlight (#1111)
* feat(pi): add native child runtime with scoped Spotlight * fix(pi): address native child review feedback * fix(pi): suppress prefaced structured child output * fix(pi): restore native child text streaming --------- Co-authored-by: wyuc <wang-yc24@mails.tsinghua.edu.cn> |
||
|
LING DUAN
|
61a7154f07 |
feat(pi): add evidence-aware Director context runtime (#971)
* feat(pi): add evidence-aware Director context runtime * fix(web-search): require completed native search * fix(orchestration): preserve whiteboard across agent handoffs * fix(pi): remove wrap-up turn bypass |
||
|
|
3eea9dc542 |
feat(runtime): complete the #869 learner-data cutover — quiz + playback onto RuntimeStore (#955)
* feat(runtime): persist quiz attempts in RuntimeStore * fix(runtime): coalesce quiz draft snapshots * fix(runtime): recover concurrent quiz attempts * fix(runtime): handle quiz completion races * fix(runtime): commit quiz review atomically * fix(runtime): dedupe concurrent quiz writes * fix(runtime): drop stale quiz drafts * fix: harden quiz runtime review recovery * fix: serialize quiz attempt identity * feat: read quiz state from runtime store * fix: persist quiz retries before resetting * fix: preserve authoritative quiz outcomes * fix: preserve legacy quiz retries during cutover * fix: reconcile legacy quiz snapshots safely * fix: drain rollover quiz write queues * fix: drain completed quiz retry queues * fix: reuse concurrent quiz retries * fix(quiz): preserve drafts across abrupt reloads * fix(quiz): recover empty retry sessions * fix(chat): abort stalled runtime state reads * fix(quiz): expose queued phases to readers * fix(quiz): retain concurrent writer tails * fix(quiz): canonicalize retry branches * fix(quiz): keep retry rollovers monotonic * fix(quiz): validate skipped retry siblings * fix(quiz): close read cutover races * test(classroom): cover legacy quiz summaries * fix(quiz): reset async consumers on scene changes * fix(quiz): close scene transition windows * test(pbl): cover launch freshness guards * fix(quiz): close cutover concurrency gaps * fix(quiz): harden retry and context freshness * fix(quiz): reject malformed legacy answers * fix(quiz): validate legacy answer values * test(quiz): cover legacy multi-answer migration * fix(merge): retire dead ChatRequestTemplate.storeState after quiz read cutover Main's three call sites built static storeState blocks that runAgentLoopFn never consumed (it always rebuilds fresh state via getStoreState); the quiz read cutover replaced that callback with the async two-phase RuntimeStore read, leaving the template field with zero consumers. Drop it. * feat(storage): conform HTTP/PG backends and reference server to RuntimeAppendOptions The quiz write path's expectedLastSeq / sessionTransition / RuntimeAppendConflictError semantics existed only in the browser backend; server-backed deployments would silently accept conflicting appends and leave completed sessions active. Forward the options over the wire, detect conflicts atomically under the PG transaction, map them to HTTP 409 RUNTIME_APPEND_CONFLICT, and rematerialize the typed error client-side so quiz retry logic works across every backend. Co-authored-by: Codex <codex@openai.com> * fix(chat): Pi single requests build storeState via the async runtime quiz read Pi bypasses runAgentLoop's per-iteration getStoreState and serializes the request template straight to /api/chat/pi, which rejects bodies without storeState. Extract the fresh-snapshot builder (async RuntimeStore quiz read with the scene-transition guard) and call it on the Pi path too. * ci: whitelist the runtime-data-cutover integration trunk for PR checks * feat(runtime): playback cutover — cursor in KV, discussion facts in RuntimeStore (#956) * feat(runtime): cut playback over to the runtime layer — cursor in KV, facts in RuntimeStore (#869) The fourth and last runtime family. Consumed-discussion facts become append-only 'playback' records folded into a set at read (at-least-once appends, no conflict machinery); the resume cursor is device-scoped last-write-wins KV per the amended #779/#869 split. sessionStorage keeps same-tab priority; KV takes over on fresh tabs/reloads. The dead Dexie playbackState machinery is retired, with a one-time lazy migration of any legacy row (cursor half + facts half) before deletion, and stage deletion now clears both the KV cursor and any unmigrated legacy row. Co-authored-by: Codex <codex@openai.com> * test(runtime): include playbackState in the stage-delete db mock --------- Co-authored-by: Codex <codex@openai.com> * fix(playback): persist discussion facts on every consumption path (#957) * fix(playback): persist discussion facts on every consumption path (final-review P0+P1s) - The engine now publishes a progress snapshot the moment a discussion is consumed (join / skip / unselected-agent auto-skip). onProgress otherwise fires before the discussion action executes and a discussion is the scene's last action, so the fact never reached persistence. - Reads fold records across ALL playback sessions in the learner partition (mergeLearner deliberately preserves same-kind sessions from both keys). - Legacy migration appends only not-yet-durable facts, so an interrupted migration resumes instead of dropping the tail. - recordConsumedDiscussion reports durability; the component drops failed ids from its observed set so a later progress tick retries (at-least-once). * test(e2e): live verification of the playback persistence chain Seeds a deterministic stage straight into the Dexie DB, starts the lecture via the canvas overlay, and asserts the full chain: discussion auto-skip appends a discussionConsumed record to maic-runtime, the device cursor lands in KV, and both survive a fresh browsing context (empty sessionStorage). * refactor(playback): consumed-discussion state is volatile by decision — cursor-only persistence (#959) Product ruling on #869's fourth family: playback learner state is front-end ephemeral UX, not learner data. A re-shown proactive card auto-skips, joined discussions' content already lives in chat runtime records, and no replay export / analytics consumer exists — so durable facts bought nothing over in-memory + same-tab sessionStorage. Drop lib/playback/runtime.ts and the RuntimeStore facts wiring; keep the device-scoped KV resume cursor (the half with real UX value), the engine's consumption-time progress snapshot (cursor freshness), and the legacy Dexie retirement (cursor half migrates, row deletes, consumed ids are dropped). * fix(review): P3 pair from cross-review — scene-id boundary + sessionTransition 400 (#966) * fix(review): scene-id boundary for quiz context + 4xx for malformed sessionTransition (P3 pair) Review findings on #955: didActiveSceneRemainUnchanged compared the active scene by object identity, so a store update reallocating the scene during the async quiz read dropped the learner's graded answers from that turn's request — the scene id is the real boundary. The records route now classifies a malformed sessionTransition as a validation failure instead of letting the store's throw surface as a 500. * fix(playback): superseded-engine cursor guard + migration write-window recheck Second-vendor review of the #959 shrink (requested after the cross-review noted it had single-vendor coverage) found: an engine orphaned by a scene switch during async lecture resume could pass the idle-only recheck, be resurrected, and publish its old scene's progress over the new scene's debounced cursor — the resume continuation now requires identity with the installed engine, and onProgress drops snapshots from superseded engines. The legacy cursor migration also rechecks KV immediately before its write so a concurrent tab's newer cursor cannot be overwritten and orphaned by the legacy-row delete. Co-authored-by: Codex <codex@openai.com> --------- Co-authored-by: Codex <codex@openai.com> * fix(review): approval follow-up P3 nits (#967) * release: v0.3.1 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(review): approval P3 nits — ISO gate on sessionTransition, dead mocks, corrupt-timestamp guard - The records route's sessionTransition guard now requires an ISO updatedAt (isIsoTimestamp), matching the sibling PATCH /status route - Dead vi.mock factories for the deleted playback-storage module dropped - A corrupt legacy playback timestamp falls back to 'now' instead of wedging migration into a permanent re-throw that disabled resume --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Codex <codex@openai.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
|
bb8d4f30e7 |
feat(chat): add experimental Pi classroom runtime behind a flag (#914)
* feat(chat): add experimental Pi classroom runtime * fix(chat): harden Pi whiteboard action sequencing * style(chat): format session retirement flow * style(playback): format async scene transition * fix(chat): retire actions on session errors * fix(chat): close Pi runtime lifecycle gaps * fix(chat): avoid replaying historical whiteboard actions * fix(chat): prevent Pi structured-output residue from leaking and inflating turns Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(chat): expose whiteboard element/line ids and bound code-line context Surface each element's id (and code line ids) in the Pi whiteboard context so later child agents can target wb_delete / wb_edit_code, and cap the rendered code lines with a shared character budget so a code-heavy board cannot grow the prompt without bound. Return only the current turn's whiteboard ledger from the Pi director loop instead of carrying history forward across requests. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(chat): bound Pi whiteboard code context Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(chat): prioritize newer persisted whiteboard code in budget Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(chat): enforce whiteboard boundaries across Pi sessions * feat(chat): refine soft-closing discussion controls * fix(chat): make whiteboard clearing model-directed --------- Co-authored-by: apple <apple@GigideMacBook-Air.local> Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> Co-authored-by: LING-6150 <duan.lin@northeastern.edu> |
||
|
danishsshaikh
|
d00795094c |
feat(playback): add action-level navigation (#843)
Squashed from #843. Adds Notes click-to-jump action navigation and action-level playback resume/restore, with visible toolbar sentence controls removed after review. |
||
|
Yanpeng Wangandwyuc
|
b6fe2814d2 |
fix(quiz): stop leaking questions on entry and pass results to chat agent (#823)
* fix(quiz): stop leaking questions on entry and pass results to chat agent Fixes #822. Two related defects on quiz scenes: 1. The `quiz-actions` opening monologue could enumerate question stems, compare options, and trigger a `discussion` action that walked agents through the quiz BEFORE the learner had answered anything. 2. After the learner submitted, the chat/discussion agent never saw their answers, correctness, the canonical `analysis`, or the `aiComment` from `/api/quiz-grade` — so post-quiz feedback was generic and often re-explained the whole quiz from scratch. Changes: - `lib/prompts/templates/quiz-actions/{system,user}.md`: collapse the generator down to a brief 1-2 segment opening, forbid `discussion` actions in scene.actions, and add explicit "no preview, no answer, no detailed concept re-teach" safety rules. - `lib/types/chat.ts` + `lib/chat/agent-loop.ts`: add optional `quizResults` to `StatelessChatRequest.storeState` / `AgentLoopStoreState` (sceneId + answers + per-question status/earned/aiComment). - `components/chat/use-chat-sessions.ts`: hydrate `quizResults` from `readSubmittedState(sceneId)` once per agent-loop iteration (and on initial dispatch) so retries are picked up. - `lib/orchestration/summarizers/state-context.ts`: when results are present, surface per-question student answer / correct answer / verdict / analysis / aiComment with a "address THIS student's mistakes" instruction; when unsubmitted, keep question visibility for clarifying questions but pin strict no-leak rules so a learner saying "I'm done" cannot bait the agent into reciting the quiz. Discussion / multi-agent round-robin and the canonical session flow are untouched. * style: apply prettier formatting to state-context --------- Co-authored-by: wyuc <wang-yc24@mails.tsinghua.edu.cn> |
||
|
wyuc
|
f064590a5e |
test+fix(orchestration): rip maxTurns; restore answer-the-user's-question behavior (#598 follow-up) (#599)
* test(orchestration): add premature-END regression eval (#598 follow-up) Adds eval/orchestration/ following the outline-language pattern: a runner that A/B-tests the director against the same scenario with two prompt variants — current main (post-#554) and a synthesised pre-#554 baseline that strips the role-aware summary labels AND the new system.md rules 10/11/12 together. Pass criterion is now framed as a regression guard rather than a fixture discrimination test: every scenario's post-fix END rate must stay below EVAL_END_THRESHOLD (default 20%). The pre-vs-post Δ is reported as informational data, since #554's reviewer feedback was that earlier fixtures didn't discriminate. Empirical finding across 7 shipped model configs (gpt-4.1-mini, gpt-4o-mini, gpt-5.4-nano, qwen-plus, qwen3.5-flash, deepseek-chat, deepseek-v4-flash, gemini-2.5-flash, claude-haiku-4-5) with 5 scenarios modelled on #511 (incl. the exact Tiananmen-3D objection trace, soft pushback after a long resolved-looking discussion, topic pivot, brief acknowledgement, and explicit teacher-signals-end-then-user-objects): no scenario produced a non-zero END rate in either variant. Useful data for the #554/#598 discussion — the prompt-layer rules don't measurably change behavior on these shipped models with these prompts, but the eval is now wired so future regressions show up. * test(orchestration): add director question-answering eval (#598 follow-up) Adds an A/B eval probing whether the director routes correctly when the conversation contains an unanswered user question — both in the first-turn drift case (agents started answering but drifted onto adjacent topics, no frustration yet) and in the escalated frustration case (user has already complained multiple times). The prod symptom this targets ("答非所问"): agents reply on adjacent topics rather than answering the literal question, and the director keeps picking peer agents for "variety" instead of routing back to the teacher to actually answer. Scenarios: 5 synthesized cases across math, biology, English grammar, physics, and calculus. 3 are first-turn drift (no frustration yet), 2 include frustration signals. Each captures: user asks specific question, agents reply on adjacent topics, director is asked to pick the next move. Decision rule-judged into USER | TEACHER | OTHER_AGENT | END: - USER → ✓ valid (cue user to clarify) - TEACHER → ✓ valid (re-route to teacher to answer) - OTHER_AGENT → ✗ wrong (peer-agent "variety" routing — the bug) - END → ✗ wrong A/B: - baseline : current director template with rule 13 stripped - with_rule : current director template as-shipped Pass = with_rule.correctRate ≥ EVAL_PASS_THRESHOLD (default 0.7). Pre-vs-post Δ is reported as informational only. Note this commit only adds the eval; the rule it strips/restores in the A/B is injected in the next commit. * fix(orchestration): inject director rule for answering user questions Adds rule 13 to lib/prompts/templates/director/system.md so the director recognizes when the conversation contains an unanswered user question and routes appropriately — to the teacher (literal role match) by default, or to USER cue when the question is ambiguous or the user has already complained multiple times. This addresses the "答非所问" symptom: when agents drift onto adjacent topics, the director keeps picking peer agents for "variety" rather than routing back to the teacher to actually answer. The rule is preventive (applies whenever the question is unanswered, including the first turn after drift) rather than reactive-only (waiting for the user to express frustration). Rule wording uses constrained-output framing: literal `role: teacher` match (not "best fit" / "highest priority" / "most knowledgeable"), and an explicit list of forbidden choices (peer agents, END). Earlier softer wordings let the model pick the assistant when it seemed more topically relevant; the literal-match wording prevents that. Validated by the question-answering eval added in the previous commit. On google:gemini-3-flash-preview, 5 samples per variant across 5 synthesized scenarios: | Scenario | baseline | with_rule | |----------------------------------------------------|----------|-----------| | math_quadratic_axis_drift_no_frustration | 100% | 100% | | bio_dark_reaction_drift_no_frustration | 80% | 100% | | english_team_isare_drift_no_frustration | 20% | 100% | | physics_inertial_mass_escalated_frustration | 0% | 100% | | calculus_product_rule_drift_no_frustration | 60% | 100% | | **mean** | **52%** | **100%** | 3 of 5 scenarios reproduce the bug at baseline (correct rate ≤ 60%); the rule lifts every scenario to 100%. The remaining 2 baselines are already high — the rule doesn't regress them. * fix(orchestration): teach agents to answer the user's literal question Adds an "Answering the User's Question" section to the agent system prompt. Before this, the agent SP had no rule for direct-answering — the persona, role guideline, current-state context, and length guidelines all pushed the agent to advance the lesson and inspire thought, with no explicit directive to first answer the user's literal question. The result observed in prod ("答非所问"): when the user asks a specific question (a formula, a yes/no, a term), agents reply with adjacent concepts, examples, or follow-up questions — but never the literal answer. The new section directs the agent, on every response where the user message contains a question, to: - lead with the concrete answer to the literal question, - not pivot to an adjacent topic even when it seems pedagogically richer, - treat "inspire thought" and peer-differentiation guidance as applying only AFTER the literal question is answered, - say "I don't know" directly when uncertain rather than answering a different question, - on frustration signals from prior turns, look back at the original question (before the frustration) and answer THAT specifically. This is the agent-side counterpart to the director rule landed in the previous commit. Director routes to the teacher when the question is unanswered; the teacher now has explicit guidance for what to do when dispatched. * refactor+fix(orchestration): rip maxTurns end-to-end; distinguish errors from completion Two coupled cleanups in one commit since the outcome-handler rewrite naturally falls out of removing the client-side max-turns loop. The director graph used `maxTurns: turnCount + 1` to force "one director→agent cycle per request" — a hack expressing the single-round contract as a turn limit. This commit removes the hack and the surrounding user-facing setting, replacing implicit intent with explicit graph topology. Server: - `director-graph.ts`: drop `maxTurns` from state, drop the hard-cap check at the top of `directorNode`, change `agent_generate → director` edge to `agent_generate → END`. Single-round contract is now expressed by topology. - `lib/types/chat.ts`: drop `maxTurns` and `currentTurn` from `SessionConfig`; drop `maxTurns?` from `CreateSessionRequest.trigger`. Client: - `lib/store/settings.ts`: drop `maxTurns` state, `setMaxTurns` action, the `'10'` default, and the localStorage migration path. - `components/agent/agent-bar.tsx`: drop the entire compact-stepper UI block plus the unused MessageSquare/Minus/Plus lucide imports. - `components/settings/agent-settings.tsx`: drop the AgentSettings prop and the multi-agent-only Input block. - `lib/chat/agent-loop.ts`: drop `maxTurns` parameter, change `while (turnCount < maxTurns)` to `while (true)` controlled by inner exits, drop `'max_turns'` reason from `AgentLoopOutcome`. - `components/chat/use-chat-sessions.ts`: drop the maxTurns lookup, drop the trailing arg to runAgentLoop, drop vestigial `maxTurns: 0` / `currentTurn: 0` from session creation. - `eval/whiteboard-layout/runner.ts`: drop unused MAX_AGENT_TURNS plus the trailing arg. i18n: drop `settings.maxTurns` and `settings.maxTurnsDesc` from six locales (en-US, zh-CN, zh-TW, ja-JP, ar-SA, ru-RU). Motivation for removing the user-facing setting: the LLM director controls round length via cue_user / END decisions, so a numeric "max discussion turns" knob is redundant with rule-driven routing. When the director prompt was buggy, the cap masked the bug by force- ending; once director routing is fixed, the cap just removes user agency. Before, every non-`cue_user` outcome of the agent loop mapped to `status: 'completed'` in `use-chat-sessions.ts`. That conflated four distinct paths: - `end`: director chose END (legitimate completion) - `empty_turns`: two consecutive agents returned no text or actions (parser failure, model timeout, etc.) - `no_done`: SSE stream ended without a `done` event (infrastructure failure) - `aborted`: user pressed stop (handled elsewhere) The new switch routes `end` to completed, `empty_turns` and `no_done` through `clearLiveSessionAfterError` so the user sees an inline System message describing what went wrong instead of a quiet "completed". - `lib/types/chat.ts`: add `'error'` to `SessionStatus`. - `components/chat/use-chat-sessions.ts`: rewrite the outcome handler as a switch; have `clearLiveSessionAfterError` set `status: 'error'` so all entry points (the runAgentLoop outcome AND existing throw paths from fetchChat / SSE error events) converge on the same visible state. - `components/chat/session-list.tsx`: add `AlertCircle` red icon for the new error status. - i18n: add `chat.error.emptyAgentResponses` and `chat.error.streamInterrupted` strings to all six locales. This is the in-product fix that the prod investigation surfaced as an independent root cause of the #511 "discussion ended unexpectedly" symptom — separate from the director and agent prompt fixes in the earlier commits of this PR. * fix(chat): drop dead settingsState binding; wire eval:orchestration:answering script Two minor CR follow-ups: - components/chat/use-chat-sessions.ts: the `settingsState` lookup at the top of runAgentLoopFn was the entry point for the maxTurns lookup ripped in the previous commit; with maxTurns gone, the binding is unused and ESLint flags it. Drop the line. (Other uses of useSettingsStore.getState() elsewhere in this file are unaffected.) - package.json: the answering eval runner had no convenience script, while the premature-end runner did. Add eval:orchestration:answering for symmetry and discoverability. * fix(orchestration): route ambiguous/unaddressed user turns to teacher, never USER cue Addresses reviewer issues 2 & 3 (ambiguous "帮我看一下这个" / "Can you take a look at this?"): the director was cueing USER for vague input, which makes no agent speak — the user faces dead air with nothing to react to ("nobody responds"); and when an agent did speak it guessed a topic instead of asking what the user meant. Root cause was the previous rule 13, which offered `{"next_agent":"USER"}` as a valid response to an ambiguous user turn. A USER cue is the wrong tool here: it returns control to the user without any agent speaking. The right behavior is to dispatch the teacher, who either answers or — if the message is too vague — asks a clarifying question out loud. Director (`director/system.md` rule 13): - When the most recent user turn is unaddressed (clear question, vague request, OR frustration signal), the only acceptable output is the literal `role: teacher` agent id. - USER cue is now explicitly forbidden as a response to a user turn (reserved for rule 7, where an agent has just asked the user something). - Tightened the END prohibition: a high turn count or a thoroughly-discussed topic does NOT mean the specific question was answered. Added rule 5 (don't-drag-on) to the override list, since that pressure was driving the director to END on well-discussed-but-unanswered turns. Agent (`agent-system/system.md`): - Added a clause to "Answering the User's Question": if the user's message is too vague to answer (no clear referent), ask ONE short clarifying question with a concrete option or two — do not guess a topic and lecture, do not stay silent. Eval (`answering-runner.ts` + `scenarios/answering.json`): - Reclassified USER as wrong (dead air); only TEACHER counts as correct now (teacher answers, or asks a clarifying question for vague input). - Added two ambiguity scenarios: a bare "帮我看一下这个" opener, and a vague "再帮我看看这个吧" follow-up after a real Q&A. - Made the rule-13 strip regex match by rule number rather than heading text, so it survives heading rewrites. On google:gemini-3-flash-preview, 10 samples per variant, all 7 scenarios PASS (with_rule mean ~97%, baseline ~57%). The two ambiguity scenarios: ambiguous_look_at_this 100%/100%, ambiguous_vague_followup 20%→100%. Manually verified the agent side: for "帮我看一下这个" the teacher now replies "…你想讨论的是具体的代码逻辑，还是 Agent 的决策流程图呢?" (clarifying question + concrete options) instead of guessing. Note: reviewer issue 1 (follow-up not answered) could not be reproduced on this branch across 7 configurations; it is consistent with the deployed environment running older code and/or provider API flakiness, which the C7 error-vs-completion distinction in this PR already surfaces instead of a silent 已结束. * docs(eval): fix stale physics scenario description after USER→wrong reclassification The physics_inertial_mass scenario description still said "USER-cue signal — but TEACHER re-answer is also valid", which contradicts the new classifier (USER is dead air = wrong; only TEACHER is correct). Cosmetic only — the description field is printed into the report, never used in classification. |
||
|
wyucand杨慎
|
e613b7578f |
[codex] add per-model thinking config (#494)
* add per-model thinking config * Refine thinking model controls --------- Co-authored-by: 杨慎 <117187635+cosarah@users.noreply.github.com> |
||
|
|
ea0e8126a5 |
feat: whiteboard layout quality eval harness (#425)
* feat(eval): add state manager bridging ActionEngine for eval Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * feat(eval): add shared types for whiteboard layout eval harness * feat(eval): add SSE chat client for whiteboard eval * feat(eval): add Playwright capture module for whiteboard screenshots * feat(eval): add VLM scorer for whiteboard layout evaluation * feat(eval): add report generator for whiteboard eval results * feat(eval): add 8 constructed scenarios for whiteboard layout eval * feat(eval): add minimal whiteboard render page for Playwright screenshots Creates app/eval/whiteboard/page.tsx — a headless client page that seeds the stageStore with a synthetic slide scene, exposes window.__setElements() for Playwright to inject PPTElement[], and renders them via ScreenElement inside a 1000×562.5px white canvas. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * feat(eval): add main runner for whiteboard layout eval Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * feat(eval): add eval:whiteboard script, install tsx, gitignore results * fix(eval): fix TS errors, lint, and prettier formatting * fix(eval): address code review — add cue_user/empty turn guards, validate VLM output, fix empty array crash * refactor(eval): replace synthetic scenarios with realistic ones - Replace 8 generic scenarios with 6 that match real usage patterns - Multi-agent discussion with short user replies (嗯, 明白了, 继续) - Include real slide scene data as initialStoreState - Generated agent configs with Chinese names and proper roles - Cover: physics, math, finance, primary school, economics, medical Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * refactor: extract shared agent loop from use-chat-sessions Extract the core agent loop logic into lib/chat/agent-loop.ts as a pure async function with callback injection. Both the frontend React hook and the eval harness now share the same loop — SSE parsing, exit conditions (END/cue_user/empty turns/max turns), and director state accumulation. The frontend wires StreamBuffer callbacks for UI pacing; the eval wires ActionEngine + message accumulation for headless execution. If loop logic changes in the shared module, both consumers automatically stay in sync. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix(eval): use project LLM infrastructure, fix eval page and model config - Rewrite scorer to use resolveModel() + generateText() from AI SDK instead of raw fetch — supports all providers (OpenAI, Google, Anthropic) - Model config via env vars (EVAL_CHAT_MODEL, EVAL_SCORER_MODEL), matching the pattern from outline-language eval - Fix eval page: bootstrap store before SceneProvider mounts - Fix __dirname for tsx CJS mode - Remove --api-key/--scorer-model CLI args (use env vars instead) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * feat(eval): organize results by model/timestamp * fix: remove double turnCount increment in shared agent loop The extracted agent-loop.ts had both `turnCount = directorState?.turnCount ?? turnCount + 1` (line 190) and a redundant `turnCount++` (line 215), causing multi-agent scenarios to hit maxTurns at half the expected number of iterations. Also removes unused processSSEStream import from use-chat-sessions.ts. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * feat(eval): organize screenshots by scenario subdirectory Results structure: results/<model>/<timestamp>/<scenario>/run0_turn1.png Report files stay at the timestamp level. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * feat(eval): revise scorer rubric and add rescore mode - Replace space_utilization with rendering_correctness and content_completeness - Rubric now evaluates from a teacher's perspective (empty space is normal) - readability emphasizes font size consistency - Add --rescore flag to re-score existing screenshots without re-running chat - Increase maxOutputTokens to 2000, add JSON parse error recovery - Score errors no longer abort the entire scenario Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * feat(eval): sharpen scorer rubric with teacher-perspective examples The rubric now catches specific classroom whiteboard failure modes: - overlap now explicitly penalizes writing over existing content when empty space is available (spatial planning failure) - rendering_correctness calls out diagram accuracy (e.g., parabola drawn as V-shape), raw subscripts (G_x), Chinese inside LaTeX math mode - content_completeness emphasizes canvas edge clipping and bare unlabeled diagrams - readability penalizes text styled as UI components (gray card backgrounds) - overall instructed to weight overlap and rendering_correctness more heavily - explicit note to ignore the "N" page UI element Also increases maxOutputTokens to 3000 since longer rubric produces longer justifications. Reporter now guards against null scores (scorer failures no longer crash report generation). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: address code review findings Critical fixes: - use-chat-sessions: restore agent_end handling, currentMessageId fallback for text_delta/action events with missing messageId, and re-throw on SSE error events (previously silently pushed to buffer only). - eval runner: serialize ActionEngine executions via promise chain. void-fire-and-forget raced with ensureWhiteboardOpen's 2s delay and could insert elements out of order or before the first element was committed to the store. Important fixes: - CHAT_MODEL default: 'openai/gpt-4o-mini' -> 'openai:gpt-4o-mini' (parseModelString splits on ':', not '/'). - CheckpointResult.score is now VlmScore | null; removed the 'as unknown as' cast that hid the null contract from consumers. - Delete dead code: eval/whiteboard-layout/chat-client.ts and components/chat/process-sse-stream.ts (both unused after the shared agent loop refactor). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Co-authored-by: 杨慎 <117187635+cosarah@users.noreply.github.com> |
||
|
wyucandClaude Opus 4.6
|
843d33e5cc |
chore: enforce Prettier formatting and fix lint issues
- Add .prettierignore to exclude vendor packages, lock files, markdown, and YAML - Update .prettierrc: printWidth 100, singleQuote, trailingComma "all" - Run Prettier across all source files for consistent formatting - Fix unused imports (UserRequirements, setTTSProvider) - Fix eslint-disable comment placement after Prettier reformat Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> |
||
|
wyucandClaude Opus 4.6
|
0d20abfed9 |
Initial commit: OpenMAIC AI classroom platform
Next.js 16 App Router application with: - AI-powered interactive classroom generation from PDF/text requirements - Multi-agent discussion system (teacher, students, assistant roles) - Provider abstraction for LLM, TTS, ASR, image, video, web search - Local-first data architecture (IndexedDB/Dexie) - i18n support (zh-CN/en-US) - Structured logger, SSRF guard, lint/build passing (0 errors) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> |