opencode-cache-hit 0.2.1 → 0.2.2

package/AGENTS.md

@@ -16,6 +16,7 @@ OpenCode TUI sidebar plugin: **cache hit rate**, **tokens**, **cost**, with **su
 | Architecture | [docs/en/design.md](docs/en/design.md) | [docs/zh-CN/design.md](docs/zh-CN/design.md) |
 | Timeline / JSONL | [docs/en/timeline.md](docs/en/timeline.md) | [docs/zh-CN/timeline.md](docs/zh-CN/timeline.md) |
 | TUI panel | [src/tui-panel/README.md](src/tui-panel/README.md) | [src/tui-panel/README.zh-CN.md](src/tui-panel/README.zh-CN.md) |
+| Migration plan | [docs/en/frontend-migration-plan.md](docs/en/frontend-migration-plan.md) | [docs/zh-CN/frontend-migration-plan.md](docs/zh-CN/frontend-migration-plan.md) |
 | Contributing / npm | [CONTRIBUTING.md](CONTRIBUTING.md) | — |
 | Index | [docs/README.md](docs/README.md) | |
 

package/docs/README.md

@@ -1,8 +1,8 @@
 # Documentation
 
-| | User guide | Design | Timeline | TUI panel |
-|--|------------|--------|----------|-----------|
-| English | [README.md](../README.md) | [en/design.md](./en/design.md) | [en/timeline.md](./en/timeline.md) | [../src/tui-panel/README.md](../src/tui-panel/README.md) |
-| 中文 | [README.zh-CN.md](../README.zh-CN.md) | [zh-CN/design.md](./zh-CN/design.md) | [zh-CN/timeline.md](./zh-CN/timeline.md) | [../src/tui-panel/README.zh-CN.md](../src/tui-panel/README.zh-CN.md) |
+| | User guide | Design | Timeline | TUI panel | Migration |
+|--|------------|--------|----------|-----------|-----------|
+| English | [README.md](../README.md) | [en/design.md](./en/design.md) | [en/timeline.md](./en/timeline.md) | [../src/tui-panel/README.md](../src/tui-panel/README.md) | [en/frontend-migration-plan.md](./en/frontend-migration-plan.md) |
+| 中文 | [README.zh-CN.md](../README.zh-CN.md) | [zh-CN/design.md](./zh-CN/design.md) | [zh-CN/timeline.md](./zh-CN/timeline.md) | [../src/tui-panel/README.zh-CN.md](../src/tui-panel/README.zh-CN.md) | [zh-CN/frontend-migration-plan.md](./zh-CN/frontend-migration-plan.md) |
 
 Contributing: [CONTRIBUTING.md](../CONTRIBUTING.md).

package/docs/en/design.md

@@ -59,7 +59,8 @@ sequenceDiagram
   Host->>API: session.list → childIds
   API-->>Host: message.updated { info: Message }
   Host->>Host: refreshTick++, timeline.handleMessage(info)
-  Host->>API: session.messages(sid / cid)
+  Host->>API: session.get(sid / cid) → aggregates
+  Host->>API: session.messages(sid / cid) → fallback / trend
   Host->>W: main, messages, subAgents
   W->>W: aggregate / format / TuiPanel
 ```
@@ -113,14 +114,14 @@ flowchart TD
   F -->|no| R
   C --> H[childSessionIdsForParent]
   C2 --> H
-  H --> I[re-read messages + aggregate]
+  H --> I["session.get aggregate + messages fallback"]
 ```
 
 - **Single source of truth**: `childIds` always **replaced** from `session.list` (no `session.get` append).
 - **Races**: `listGen` increments on parent change; callbacks check generation and `parentId`.
 - **Streaming**: foreign-session `message.updated` is debounced (`CHILD_LIST_DEBOUNCE_MS` = 200ms).
 - **Depth**: direct children only (`parentID === sid`); nested sub-agents are future work.
-- **Data**: `messages(cid)` → `aggregateSessionFromMessages`; entries without stats are filtered out.
+- **Data**: `session.get(cid)` → `aggregateFromSessionObject` (primary); falls back to `messages(cid)` → `aggregateSessionFromMessages`. Model/provider metadata is patched from messages when missing from session aggregates (`withModelFallback`).
 
 ### Agents block semantics
 
@@ -154,12 +155,12 @@ Implementation: `agents-view.tsx` calls `formatSubAgentLabel` + `modelRowColor`;
 
 | Data | Trigger |
 |------|---------|
-| Main snapshot | `createMemo` reads `refreshTick` + `messages(sid)` |
-| Main messages (Hit trend) | Same |
-| Sub-agent content | `childIds` or `refreshTick` → re-read each `messages(cid)` |
+| Main snapshot | `createMemo` reads `refreshTick` + `session.get(sid)` (aggregate); fallback `messages(sid)` |
+| Main messages (Hit trend) | `createMemo` reads `refreshTick` + `messages(sid)` |
+| Sub-agent content | `refreshTick` + `childIds` → `session.get(cid)` per child; fallback `messages(cid)` |
 | Sub-agent ids | `session.list` callback / debounced refresh on foreign activity |
 
-`sidebar-host` subscribes to `message.updated` and bumps `refreshTick` so streaming token updates always recompute.
+`sidebar-host` subscribes to `message.updated` and bumps `refreshTick` so dependent memos recompute; session totals follow `session.get()` aggregates, while per-turn trend follows recent messages.
 
 ### Accumulation rules
 

package/docs/en/frontend-migration-plan.md

@@ -0,0 +1,100 @@
+# Frontend Migration: session.get() Aggregates
+
+## Overview
+
+Cache-hit previously derived all cost/token statistics by iterating `api.state.session.messages()` and summing per-message fields. This path is capped at **100 messages** by the OpenCode TUI sync layer ([issue #31513]), silently truncating sessions with >100 messages.
+
+**Impact**: For a session exceeding 100 messages, the cap causes significant data loss. In a real long-running session, the plugin showed only ~43% of actual cache-read tokens and ~39% of actual cost.
+
+| Metric | Old (messages, 100-cap) | New (session.get, no cap) |
+|--------|------------------------|--------------------------|
+| Cache read | 59.6M tok | **139.0M tok** |
+| Cache write | 89.8K tok | **1.4M tok** |
+| Miss (input) | 118 tok | **329 tok** |
+| Output | 32.7K tok | **91.9K tok** |
+| Sub-agents visible | 3 | **10** |
+
+The fix swaps the data source from per-message iteration to `api.state.session.get()`, which returns **database-level aggregates** computed from all messages by the session projector — not affected by the 100-message cap.
+
+## Root Cause
+
+OpenCode's TUI sync (`packages/opencode/src/cli/cmd/tui/context/sync.tsx`) hardcodes `limit: 100` when fetching session messages:
+
+```typescript
+// Line 559 — fetches at most 100 messages
+sdk.client.session.messages({ sessionID, limit: 100 }),
+
+// Lines 580-581 — TUI store also keeps only last 100
+const visible = infos.slice(-100)
+```
+
+`api.state.session.messages(sid)` reads from `sync.data.message[sid]`, which holds at most 100 entries. For sessions exceeding this, plugins see incomplete data with no indication of truncation.
+
+Meanwhile, the `session` table in OpenCode's SQLite database stores **pre-aggregated** columns (`cost`, `tokens_input`, `tokens_output`, `tokens_cache_read`, `tokens_cache_write`) updated via SQL increments by the session projector — every message, no cap.
+
+`api.state.session.get(sid)` returns the full `Session` object (SDK type `Session`), which includes `cost` and `tokens` fields populated from these database aggregates.
+
+## Solution
+
+### Design
+
+```
+Before:
+  api.state.session.messages(sid) → iterate 100 msgs → sum           ❌ capped
+
+After:
+  api.state.session.get(sid)      → read cost/tokens directly → done  ✅ full
+  └─ session not available?       → fallback to messages             (compat)
+```
+
+### Changes
+
+| File | Change |
+|------|--------|
+| `src/types.ts` | New `SessionObject` type (`model`, `cost`, `tokens`, `parentID`); updated `session.get` return type |
+| `src/stats.ts` | New `aggregateFromSessionObject()` — O(1) extraction from `SessionObject` → `SessionSnapshot` |
+| `src/sidebar-host.tsx` | `mainSnap` and `subAgentList` preferred `session.get()` with fallback to message iteration |
+| `tests/stats.test.ts` | 3 test cases covering empty, full, and partial session objects |
+
+### Data Flow
+
+1. **`mainSnap`**: Calls `api.state.session.get(sid)` → `aggregateFromSessionObject()` → returns snapshot if stats present. If session unavailable or empty, falls back to `aggregateSessionFromMessages()`.
+
+2. **`subAgentList`**: Same pattern per child session ID.
+
+3. **Per-call trend** (`computePerCallHitTrend`): Still uses `api.state.session.messages()` — trend only needs the last few turns, so 100 messages is sufficient.
+
+### Why session.get() works for resumed sessions
+
+When resuming a session, OpenCode's TUI `sync()` immediately calls `sdk.client.session.get({ sessionID })` which returns the full `Session` object with database-level aggregates. This object is stored in the TUI state store, and `api.state.session.get(sid)` reads it. The plugin sees accurate totals from the moment the session loads.
+
+### Fallback rationale
+
+`session.get()` returns `undefined` for sessions that have never been synced to the TUI state (rare, but possible for child sessions loaded in a different context). The fallback to `api.state.session.messages()` preserves existing behavior as a safety net.
+
+## Performance
+
+| Metric | Old (message iteration) | New (session.get) |
+|--------|------------------------|-------------------|
+| Time complexity | O(n) where n ≤ 100 | O(1) |
+| Memory | Reads entire message array | Reads one small object |
+| Reactivity trigger | `message.updated` → recompute | Same event, same timing |
+| Data completeness | ≤ 100 messages | All messages (DB aggregate) |
+
+## Rollout
+
+1. **No config changes required** — the plugin automatically uses `session.get()` when available.
+2. **No migration step** for existing users — old config files and timeline logs are unaffected.
+3. **Restart OpenCode** to pick up the updated plugin code.
+4. **Verify**: Open a session with >100 messages; the sidebar should show totals matching database aggregates.
+
+## Related
+
+- [anomalyco/opencode#31513] — v1 TUI sync hardcodes limit:100
+- [anomalyco/opencode#26861] — PR adding cursor-based pagination (pending)
+- [anomalyco/opencode#6548] — Paginated message loading feature request
+
+[issue #31513]: https://github.com/anomalyco/opencode/issues/31513
+[anomalyco/opencode#31513]: https://github.com/anomalyco/opencode/issues/31513
+[anomalyco/opencode#26861]: https://github.com/anomalyco/opencode/pull/26861
+[anomalyco/opencode#6548]: https://github.com/anomalyco/opencode/issues/6548

package/docs/en/timeline-duplicate-writes.md

@@ -12,10 +12,10 @@ The OpenCode TUI plugin API has two data access patterns:
 
 | API | Behavior |
 |-----|----------|
-| `api.state.session.messages(sessionId)` | Returns **all** messages — no cursor, `since`, or limit |
+| `api.state.session.messages(sessionId)` | Returns up to 100 messages (capped by TUI sync); no cursor, `since`, or limit parameter on the plugin side |
 | `api.event.on("message.updated", handler)` | Fires per-message, carries the **full `Message` object** (cost, tokens, time, modelID, providerID) — see [`types.gen.ts` L1064](https://github.com/anomalyco/opencode/blob/dev/packages/sdk/js/src/v2/gen/types.gen.ts#L1064) |
 
-The original collector used the **polling path**: subscribe to `message.updated` → call `schedule()` (500ms debounce) → `collectNow()` → `getMessages()` (full history) → `buildCallRecords()` (all messages) → `shouldFlushToDisk()` (dedup filter). This required a `flushedKeys` Set to skip already-written records.
+The original collector used the **polling path**: subscribe to `message.updated` → call `schedule()` (500ms debounce) → `collectNow()` → `getMessages()` (TUI-synced window, up to 100 messages) → `buildCallRecords()` (all returned messages) → `shouldFlushToDisk()` (dedup filter). This required a `flushedKeys` Set to skip already-written records.
 
 On restart, `flushedKeys` was lost (it lived in memory only). Every message in the session would be re-flushed to JSONL. A startup scan of all JSONL files was proposed to rebuild the set, adding complexity without addressing the core issue: **polling is the wrong pattern for a write-only log**.
 

package/docs/zh-CN/design.md

@@ -59,7 +59,8 @@ sequenceDiagram
   Host->>API: session.list → childIds
   API-->>Host: message.updated { info: Message }
   Host->>Host: refreshTick++, timeline.handleMessage(info)
-  Host->>API: session.messages(sid / cid)
+  Host->>API: session.get(sid / cid) → 聚合值
+  Host->>API: session.messages(sid / cid) → fallback / 趋势
   Host->>W: main, messages, subAgents
   W->>W: aggregate / format / TuiPanel
 ```
@@ -113,14 +114,14 @@ flowchart TD
   F -->|否| R
   C --> H[childSessionIdsForParent]
   C2 --> H
-  H --> I[messages 重读 + 聚合]
+  H --> I["session.get 聚合 + messages 兜底"]
 ```
 
 - **唯一来源**：`childIds` 始终由 `session.list` 结果**覆盖**写入，不再 `session.get` + 追加（避免漏发现、僵尸 id）。
 - **竞态**：`listGen` 在 parent 切换时递增；回调校验 generation 与 `parentId` 未变。
 - **流式**：外国 session 的 `message.updated` 很密，用 `CHILD_LIST_DEBOUNCE_MS`（200ms）合并 list 请求。
 - **一层子 session**：`parentID === sid` 的直接子节点；嵌套子 agent 见「未来方向」。
-- **子 session 数据**：对每个 `cid` 调用 `messages(cid)` → `aggregateSessionFromMessages`；无统计的条目过滤掉。
+- **子 session 数据**：`session.get(cid)` → `aggregateFromSessionObject`（主路径）；不可用时降级到 `messages(cid)` → `aggregateSessionFromMessages`。若 session 聚合缺少 model/provider 元数据，从 messages 末尾补齐（`withModelFallback`）。
 
 ### Agents 合计语义（实现正确，勿与「全场总账」混淆）
 
@@ -154,12 +155,12 @@ flowchart TD
 
 | 数据 | 触发方式 |
 |------|----------|
-| 主 session snapshot | `createMemo` 内读 `refreshTick` + `api.state.session.messages(sid)` |
-| 主 session 消息列表（Hit 趋势） | 同上 |
-| 子 agent 列表内容 | `childIds` 变化或 `refreshTick` 后各 `messages(cid)` 重读 |
+| 主 session snapshot | `createMemo` 内读 `refreshTick` + `session.get(sid)`（聚合）；fallback `messages(sid)` |
+| 主 session 消息列表（Hit 趋势） | `createMemo` 内读 `refreshTick` + `messages(sid)` |
+| 子 agent 列表内容 | `refreshTick` + `childIds` → 每个 child 的 `session.get(cid)`；fallback `messages(cid)` |
 | 子 agent id 集合 | `session.list` 完成回调 / `message.updated` 发现新 child |
 
-主 session **显式**订阅 `message.updated`（在 `sidebar-host`）：每次事件 `refreshTick++`，保证流式过程中 `tokens` 更新会重算（不仅依赖 store 是否自动触发 Solid memo）。
+主 session **显式**订阅 `message.updated`（在 `sidebar-host`）：每次事件 `refreshTick++`，触发相关 memo 重算；会话总量以 `session.get()` 聚合为准，逐轮趋势使用最近 messages。
 
 ### 累加规则
 

package/docs/zh-CN/frontend-migration-plan.md

@@ -0,0 +1,100 @@
+# 前端迁移方案：session.get() 聚合数据
+
+## 概述
+
+cache-hit 插件此前通过遍历 `api.state.session.messages()` 逐条累加 per-message 字段来统计 cost/token。这个路径受 OpenCode TUI sync 的 **100 条消息上限** 限制（[issue #31513]），超过 100 条消息的会话数据被静默截断。
+
+**影响**：对于超过 100 条消息的会话，截断导致大量数据丢失。在一个实际的长会话中，插件仅显示了约 43% 的缓存读 token 和约 39% 的实际费用。
+
+| 指标 | 旧（messages，100 条截断） | 新（session.get，无截断） |
+|------|--------------------------|-------------------------|
+| 缓存读 | 59.6M tok | **139.0M tok** |
+| 缓存写 | 89.8K tok | **1.4M tok** |
+| 未命中 | 118 tok | **329 tok** |
+| 输出 | 32.7K tok | **91.9K tok** |
+| 可见子 Agent | 3 | **10** |
+
+修复方案是将数据来源从逐条消息累加切换为 `api.state.session.get()`，该接口返回 **数据库层聚合值**，由 session projector 从全部消息累加得出 —— 不受 100 条上限影响。
+
+## 根因
+
+OpenCode TUI sync（`packages/opencode/src/cli/cmd/tui/context/sync.tsx`）在获取会话消息时硬编码了 `limit: 100`：
+
+```typescript
+// 第 559 行 — 最多获取 100 条消息
+sdk.client.session.messages({ sessionID, limit: 100 }),
+
+// 第 580-581 行 — TUI store 也只保留最后 100 条
+const visible = infos.slice(-100)
+```
+
+`api.state.session.messages(sid)` 读取的是 `sync.data.message[sid]`，最多 100 条。超过的会话，插件看到的是不完整数据，且没有任何截断提示。
+
+与此同时，OpenCode SQLite 数据库的 `session` 表存储了**预聚合**列（`cost`、`tokens_input`、`tokens_output`、`tokens_cache_read`、`tokens_cache_write`），由 session projector 通过 SQL 增量更新 —— 每条消息都会被计入，无上限。
+
+`api.state.session.get(sid)` 返回完整 `Session` 对象（SDK 类型 `Session`），包含来自数据库聚合的 `cost` 和 `tokens` 字段。
+
+## 方案
+
+### 设计
+
+```
+之前：
+  api.state.session.messages(sid) → 遍历 100 条消息 → 累加           ❌ 截断
+
+现在：
+  api.state.session.get(sid)      → 直接读 cost/tokens → 完成       ✅ 全量
+  └─ session 不可用时              → fallback 到消息累加             （兼容）
+```
+
+### 改动
+
+| 文件 | 内容 |
+|------|------|
+| `src/types.ts` | 新增 `SessionObject` 类型（`model`、`cost`、`tokens`、`parentID`）；更新 `session.get` 返回类型 |
+| `src/stats.ts` | 新增 `aggregateFromSessionObject()` — O(1) 从 `SessionObject` 提取 `SessionSnapshot` |
+| `src/sidebar-host.tsx` | `mainSnap` 和 `subAgentList` 优先使用 `session.get()`，不可用时 fallback 到消息累加 |
+| `tests/stats.test.ts` | 3 个测试用例覆盖空对象、完整对象、部分字段对象 |
+
+### 数据流
+
+1. **`mainSnap`**：调用 `api.state.session.get(sid)` → `aggregateFromSessionObject()` → 有数据则返回。若 session 不可用或为空，fallback 到 `aggregateSessionFromMessages()`。
+
+2. **`subAgentList`**：每个子会话同上。
+
+3. **逐轮趋势**（`computePerCallHitTrend`）：仍使用 `api.state.session.messages()` —— 趋势只看最近几轮，100 条足够。
+
+### 为什么 resume 会话立即可见
+
+resume 会话时，OpenCode TUI `sync()` 立即调用 `sdk.client.session.get({ sessionID })` 获取完整 `Session` 对象（含数据库级聚合值）。该对象存入 TUI state store，`api.state.session.get(sid)` 读取之。插件从会话加载的瞬间就能看到准确总数。
+
+### Fallback 策略
+
+`session.get()` 对从未被 sync 到 TUI state 的会话返回 `undefined`（极少见，多见于特殊上下文中加载的子会话）。fallback 到 `api.state.session.messages()` 作为安全网保留原有行为。
+
+## 性能
+
+| 维度 | 旧（消息遍历） | 新（session.get） |
+|------|---------------|-------------------|
+| 时间复杂度 | O(n)，n ≤ 100 | O(1) |
+| 内存 | 读取整个消息数组 | 读取一个小对象 |
+| 响应式触发 | `message.updated` → 重算 | 同一事件，同一时机 |
+| 数据完整性 | ≤ 100 条消息 | 全部消息（数据库聚合） |
+
+## 上线
+
+1. **无需配置变更** —— 插件自动优先使用 `session.get()`。
+2. **存量用户无需迁移** —— 旧配置文件和 timeline 日志不受影响。
+3. **重启 OpenCode** 加载更新后的插件代码。
+4. **验证**：打开一个超过 100 条消息的会话，侧边栏数据应与数据库聚合值一致。
+
+## 相关链接
+
+- [anomalyco/opencode#31513] — v1 TUI sync 硬编码 limit:100
+- [anomalyco/opencode#26861] — 基于游标的分页 PR（待合入）
+- [anomalyco/opencode#6548] — 分页消息加载功能请求
+
+[issue #31513]: https://github.com/anomalyco/opencode/issues/31513
+[anomalyco/opencode#31513]: https://github.com/anomalyco/opencode/issues/31513
+[anomalyco/opencode#26861]: https://github.com/anomalyco/opencode/pull/26861
+[anomalyco/opencode#6548]: https://github.com/anomalyco/opencode/issues/6548

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-cache-hit",
-  "version": "0.2.1",
+  "version": "0.2.2",
   "description": "OpenCode TUI sidebar: prompt cache hit rate, tokens & cost with sub-agent rollup. Works with opencode-visual-cache; optional per-call JSONL timeline.",
   "type": "module",
   "license": "MIT",

package/src/sidebar-host.tsx

@@ -6,16 +6,20 @@ import { createTimelineCollector } from "./timeline/collector.ts"
 import type { AssistantMessage, OpenCodeTuiApi, SubAgentSummary } from "./types.ts"
 import {
   emptySessionSnapshot,
+  aggregateFromSessionObject,
   aggregateSessionFromMessages,
+  mainSessionHasStats,
   subAgentHasStats,
   toSubAgentSummary,
+  withModelFallback,
 } from "./stats.ts"
 import { createChildSessionSync } from "./child-session-sync.ts"
 import { loadPluginConfig } from "./load-config.ts"
 
 /**
- * Session-scoped sidebar host. Bumps `refreshTick` on message.updated (same as visual-cache)
- * so memos re-read api.state.session.messages.
+ * Session-scoped sidebar host. Bumps `refreshTick` on message.updated
+ * so memos re-compute. Prefers session.get() for aggregate cost/tokens
+ * (DB-level, not capped at 100 messages), falls back to session.messages().
  * Timeline writes are event-driven: message.updated → handleMessage → appendFile.
  */
 export function CacheHitSidebarHost(props: {
@@ -63,6 +67,14 @@ export function CacheHitSidebarHost(props: {
     void refreshTick()
     const sid = props.sessionId
     if (!sid) return emptySessionSnapshot()
+    const session = props.api.state.session.get(sid)
+    if (session) {
+      const snap = aggregateFromSessionObject(session)
+      if (mainSessionHasStats(snap)) {
+        const msgs = props.api.state.session.messages(sid) as AssistantMessage[] | undefined
+        return msgs ? withModelFallback(snap, msgs) : snap
+      }
+    }
     const msgs = props.api.state.session.messages(sid)
     return msgs?.length
       ? aggregateSessionFromMessages(msgs as AssistantMessage[])
@@ -76,17 +88,27 @@ export function CacheHitSidebarHost(props: {
     return (props.api.state.session.messages(sid) ?? []) as AssistantMessage[]
   })
 
-  const subAgentList = createMemo(() =>
-    childIds()
+  const subAgentList = createMemo(() => {
+    void refreshTick()
+    return childIds()
       .map((cid) => {
+        const session = props.api.state.session.get(cid)
+        if (session) {
+          const snap = aggregateFromSessionObject(session)
+          if (subAgentHasStats(snap)) {
+            const msgs = props.api.state.session.messages(cid) as AssistantMessage[] | undefined
+            const merged = msgs ? withModelFallback(snap, msgs) : snap
+            return toSubAgentSummary(cid, merged)
+          }
+        }
         const msgs = props.api.state.session.messages(cid)
         if (!msgs?.length) return null
         const snap = aggregateSessionFromMessages(msgs as AssistantMessage[])
         if (!subAgentHasStats(snap)) return null
         return toSubAgentSummary(cid, snap)
       })
-      .filter(Boolean) as SubAgentSummary[],
-  )
+      .filter(Boolean) as SubAgentSummary[]
+  })
 
   createEffect(() => {
     const sid = props.sessionId

package/src/stats.ts

@@ -1,4 +1,4 @@
-import type { AssistantMessage, SessionSnapshot, SubAgentSummary } from "./types.ts"
+import type { AssistantMessage, SessionObject, SessionSnapshot, SubAgentSummary } from "./types.ts"
 
 export function mainSessionHasStats(main: SessionSnapshot): boolean {
   return (
@@ -14,6 +14,21 @@ export function emptySessionSnapshot(): SessionSnapshot {
   return { model: "", providerID: "", input: 0, output: 0, reasoning: 0, cacheRead: 0, cacheWrite: 0, cost: 0 }
 }
 
+export function aggregateFromSessionObject(session: SessionObject): SessionSnapshot {
+  const t = session.tokens
+  const c = t?.cache
+  return {
+    model: session.model?.id ?? "",
+    providerID: session.model?.providerID ?? "",
+    input: t?.input ?? 0,
+    output: t?.output ?? 0,
+    reasoning: t?.reasoning ?? 0,
+    cacheRead: c?.read ?? 0,
+    cacheWrite: c?.write ?? 0,
+    cost: session.cost ?? 0,
+  }
+}
+
 export function aggregateSessionFromMessages(messages: readonly AssistantMessage[]): SessionSnapshot {
   let model = "",
     providerID = "",
@@ -81,6 +96,31 @@ export function subAgentHasStats(snap: SessionSnapshot): boolean {
   )
 }
 
+/**
+ * Fill missing model / providerID from the last assistant message.
+ * Session aggregates may have cost/tokens but lack model metadata;
+ * this avoids losing pricing/display when session.get() is used.
+ */
+export function withModelFallback(
+  snap: SessionSnapshot,
+  messages: readonly AssistantMessage[],
+): SessionSnapshot {
+  if (snap.model && snap.providerID) return snap
+
+  let model = snap.model
+  let providerID = snap.providerID
+  for (let i = messages.length - 1; i >= 0; i--) {
+    const m = messages[i]
+    if (m.role !== "assistant") continue
+    if (!model && m.modelID) model = m.modelID
+    if (!providerID && m.providerID) providerID = m.providerID
+    if (model && providerID) break
+  }
+  return model === snap.model && providerID === snap.providerID
+    ? snap
+    : { ...snap, model, providerID }
+}
+
 export function sidebarShouldShow(
   main: SessionSnapshot,
   subs: readonly SubAgentSummary[],

package/src/types.ts

@@ -53,13 +53,26 @@ export type ProviderInfo = {
   models: { [key: string]: { cost: ModelCost } }
 }
 
+/** Session aggregate from `api.state.session.get()` — DB-level totals, not capped by message limit. */
+export type SessionObject = {
+  model?: { id: string; providerID: string }
+  cost?: number
+  tokens?: {
+    input?: number
+    output?: number
+    reasoning?: number
+    cache?: { read?: number; write?: number }
+  }
+  parentID?: string
+}
+
 export type OpenCodeTuiApi = {
   state: {
     path: { directory: string }
     provider: ReadonlyArray<ProviderInfo>
     session: {
       messages: (id: string) => unknown[] | undefined
-      get: (id: string) => { parentID?: string } | undefined
+      get: (id: string) => SessionObject | undefined
     }
   }
   client: {