[diyor28]

Что нового для пользователя

Архитектура «движок + интерфейс»

Вся логика дебатов вынесена в отдельный модуль-движок. Интерфейс (TUI) теперь лишь отображает то, что движок решает. Это значит, что в будущем можно будет подключить веб-интерфейс, REST API или запуск без экрана — вся логика останется одной и той же.

Живая лента активности во время интервью с Opus

Раньше, пока Opus формулировал вопросы или дорабатывал тему, экран был практически пустым — только строка статуса и тусклые логи. Теперь:

• Видно, что происходит — в ленте появляются записи «Opus думает…» со спиннером, а после ответа — краткая выдержка из того, что он написал.
• Стриминг в реальном времени — пока Opus генерирует ответ (30–120 сек), вы видите текст по мере его появления, а не ждёте пустой экран.
• Прогресс по шагам — в заголовке показывается «Discovery — Turn 2/5», чтобы было понятно, сколько шагов осталось до начала дебатов.

Вопросы Opus не закрывают экран

Раньше, когда Opus задавал уточняющие вопросы, вся лента исчезала и оставалась только форма ответа. Теперь экран делится пополам: сверху — лента с контекстом, снизу — вопросы. Вы видите, о чём шла речь, пока отвечаете. На маленьких терминалах всё по-прежнему показывается в полный экран.

Системные события в ленте

Теперь в ленте явно отмечается:

• когда Opus задал вопросы и на каком шаге
• когда тема была обновлена (и сколько символов)
• когда бриф зафиксирован и дебаты начинаются

[coderabbitai]

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

• @coderabbitai resume to resume automatic reviews.
• @coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

• ▶️ Resume reviews
• 🔍 Trigger review

Walkthrough

This pull request introduces a comprehensive Engine-based state machine for debate workflows. It adds abstractions for debate formats, prompt building, session persistence, tools, actions, and lifecycle states. The main entry point is refactored to use Engine-based API access instead of direct field manipulation.

Changes

Cohort / File(s)	Summary
Application Entry & Dependencies cmd/warroom/main.go, go.mod	Refactors API access to use model.Engine.API() and model.Engine.SetAPI() instead of direct field assignment; updates Go toolchain to 1.24.7 and removes indirect testify dependency.
Engine State & Phases internal/engine/state.go, internal/engine/action.go	Introduces Phase type with 13 constants (init, api_key_setup, topic_input, etc.), Activity struct for event logging, and Action interface with 11 concrete action types (CallAPIAction, CallDiscoveryAction, BatchAction, ErrorAction, etc.) for asynchronous work.
Engine Core & Lifecycle internal/engine/engine.go	Adds comprehensive 874-line Engine struct implementing Debate interface with debate state machine, lifecycle handlers (Init, HandleAPIResult, HandleDiscoveryResult, HandleTopicLoaded, etc.), discovery flow, topic updates, and extensive logging and activity tracking.
Format & Prompt Abstractions internal/engine/format.go, internal/engine/prompt.go	Introduces Format interface and DefaultFormat implementation for debate structure/model registry; adds PromptBuilder interface with DefaultPromptBuilder for round-specific system/user prompts, research dossiers, and prior content synthesis; includes helpers (BuildResearchDossier, BuildPriorRoundContent, etc.).
Session & Tool Management internal/engine/session.go, internal/engine/tools.go, internal/engine/context.go	Adds SessionWriter interface with DefaultSessionWriter for filesystem-backed persistence; ToolRegistry interface with DefaultToolRegistry for tool definitions and execution; ContextGatherer interface and ContextItem struct for contextual data gathering.
Engine Tests internal/engine/engine_test.go, internal/engine/format_test.go, internal/engine/prompt_helpers_test.go	Comprehensive test coverage (754 lines for engine_test.go) verifying initialization paths, topic handling, discovery flow, round progression, session resume, state snapshots, cost tracking, activity lifecycle, and action dependency propagation; tests for format and prompt helpers.
TUI Layer Updates internal/tui/layout.go, internal/tui/model_test.go	Replaces config imports with engine imports; changes roundStatusIndicator signature from string phase to engine.Phase; updates model color lookup via DefaultFormat.ModelColor; refactors Model tests to use Engine-centric API (Engine.API(), Engine.State()) and new action types.
Test Snapshots internal/sim/testdata/snapshots/init.golden	Updates elapsed time in snapshot header from 0s to 1s; all other content remains identical.

Sequence Diagrams

sequenceDiagram
    participant Client
    participant Engine
    participant API as OpenRouter API
    participant SessionWriter
    participant PromptBuilder

    Client->>Engine: Init(hasAPIKey)
    alt No API Key
        Engine->>Client: Return APIKeySetupAction
    else API Key Present
        alt No Topic
            Engine->>Client: Return TopicInputAction
        else Topic Set
            Engine->>PromptBuilder: SystemPrompt(model, round, context)
            PromptBuilder-->>Engine: Prompt text
            Engine->>Client: Return CallAPIAction
            Client->>API: Send request with prompt
            API-->>Client: Response + tokens
            Client->>Engine: HandleAPIResult(content, tokens, error)
            Engine->>SessionWriter: WriteRound(results)
            SessionWriter-->>Engine: Success
            Engine->>Client: Return next action
        end
    end

Loading

sequenceDiagram
    participant Client
    participant Engine
    participant API as OpenRouter API
    participant Tools as ToolRegistry
    participant SessionWriter

    Client->>Engine: HandleTopicInput(topic, withDiscovery=true)
    Engine->>Engine: Phase = discovery
    Engine->>API: Send discovery prompt
    API-->>Engine: Tool calls + content
    Engine->>Engine: HandleDiscoveryResult(content, toolCalls)
    alt Tool = update_topic
        Engine->>SessionWriter: Update topic
        Engine->>Engine: Continue discovery loop
        Engine->>Client: Return CallDiscoveryAction
    else Tool = ask_user
        Engine->>Client: Return AskUserAction
    else No tools
        Engine->>Engine: Phase = smoke_test
        Engine->>Client: Return CallAPIAction (smoke test)
    end

Loading

sequenceDiagram
    participant Client
    participant Engine
    participant API as OpenRouter API
    participant PromptBuilder
    participant SessionWriter

    loop For each Round
        loop For each Model in Round
            Client->>Engine: HandleAPIResult(model, content, tokens)
            Engine->>SessionWriter: WriteRoundModel(result)
            Engine->>Engine: Move to next model
            Engine->>PromptBuilder: SystemPrompt(nextModel, round, context)
            Engine->>Client: Return CallAPIAction
        end
        Client->>Engine: HandleSummaryDone(roundSummary)
        Engine->>SessionWriter: WriteRoundSummary(summary)
        Client->>Engine: HandleAutoModDone(autoModNote)
        Engine->>SessionWriter: WriteRound (finalized)
        alt All rounds complete
            Engine->>SessionWriter: WriteVerdict(result)
            Engine->>Client: Return complete
        else More rounds
            Engine->>Engine: Advance to next round
            Engine->>Client: Return next CallAPIAction
        end
    end

Loading

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~60 minutes

Poem

🐰 A state machine hops through debate,
Formats and prompts align their fate,
Discovery tunnels, rounds take their turn,
Sessions persist, and lessons we learn,
The Engine runs swift—no need to wait!

---

Note

🎁 Summarized by CodeRabbit Free

Your organization is on the Free plan. CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please upgrade your subscription to CodeRabbit Pro by visiting https://app.coderabbit.ai/login.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[Copilot]

Pull request overview

Refactors the app by extracting debate orchestration into a new internal/engine synchronous state machine, making the TUI a thin adapter and enabling additional frontends (web/headless) to share the same core logic.

Changes:

• Added internal/engine package with Engine, Action types, format/prompt/tool/session abstraction interfaces, and extensive tests.
• Updated TUI rendering/tests to consume engine phases/orders and refreshed simulation golden snapshots accordingly.
• Improved OpenRouter transport body close error handling and adjusted main to use a setter for swapping APIs.

Reviewed changes

Copilot reviewed 25 out of 25 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
internal/tui/view.go	Updates roster/stats rendering to align with new engine/config usage.
internal/tui/model.go	Adds SessionLoader override plumbing and introduces SetAPI for main/cassettes.
internal/tui/layout.go	Switches phase typing for round indicator and sources model colors via engine format.
internal/tui/model_test.go	Updates tests to work with engine phases and adds an engine prompt override test.
internal/sim/testdata/snapshots/smoke_test.golden	Snapshot updated for new roster ordering/rendering.
internal/sim/testdata/snapshots/round_summary.golden	Snapshot updated for new roster ordering/rendering.
internal/sim/testdata/snapshots/round_automod.golden	Snapshot updated for new roster ordering/rendering.
internal/sim/testdata/snapshots/round.golden	Snapshot updated for new roster ordering/rendering.
internal/sim/testdata/snapshots/loading_topic.golden	Snapshot updated for new roster ordering/rendering.
internal/sim/testdata/snapshots/init.golden	Snapshot updated for new roster ordering/rendering.
internal/sim/testdata/snapshots/complete.golden	Snapshot updated for new roster ordering/rendering.
internal/openrouter/client.go	Preserves body-close errors when request body read succeeds.
internal/engine/tools.go	Introduces ToolRegistry abstraction and default registry wrapper.
internal/engine/state.go	Adds engine Phase and structured Activity events.
internal/engine/session.go	Adds SessionWriter abstraction with default filesystem implementation.
internal/engine/prompt_helpers_test.go	Moves/updates prompt helper tests into the engine package and expands cases.
internal/engine/prompt.go	Adds PromptBuilder + default implementation and migrates prompt context builders.
internal/engine/format_test.go	Adds tests for the default debate format and model registry behavior.
internal/engine/format.go	Adds Format/DebateStructure/ModelRegistry abstractions and DefaultFormat.
internal/engine/engine_test.go	Adds broad engine state machine test suite.
internal/engine/engine.go	Implements the engine state machine, actions, discovery loop, and persistence hooks.
internal/engine/context.go	Adds forward-looking ContextGatherer interface.
internal/engine/action.go	Adds typed Action values for frontends to execute async work.
go.mod	Removes an indirect testify requirement entry (dependency is still used directly elsewhere).
cmd/warroom/main.go	Swaps direct API field mutation for Model.SetAPI.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[takhir-iota]

LGTM!