fix: refine token usage reporting

The old token UI mixed context pressure, cache counters, and dollar estimates.

That made the percentage look precise even when provider data was incomplete or different.

The composer and /cost view now show concrete counts instead of a pie percentage.

Token payloads now share a smaller shape: used, inputTokens, outputTokens, and breakdown.

Claude uses per-step usage where available and Codex reads total_token_usage events.

Gemini reads its tokens object without inventing a context window.

OpenCode reads opencode.db session totals and includes all token columns in used.

The /cost backend no longer returns cache display fields or input/output dollar estimates.

This avoids derived values that look reliable but are not comparable across providers.

Verification: npm run typecheck; targeted eslint; OpenCode session provider test.

server/modules/providers/list/opencode/opencode-sessions.provider.ts

@@ -28,9 +28,9 @@ type OpenCodeHistoryRow = {
 type OpenCodeTokenTotals = {
   inputTokens: number;
   outputTokens: number;
-  cacheReadTokens: number;
-  cacheCreationTokens: number;
   reasoningTokens: number;
+  cacheReadTokens: number;
+  cacheWriteTokens: number;
 };
 
 const openOpenCodeDatabase = (): Database.Database | null => {
@@ -106,11 +106,13 @@ const buildTokenUsage = (totals: OpenCodeTokenTotals | undefined): AnyRecord | u
   }
 
   const inputTokens = totals.inputTokens;
+  const displayInputTokens = inputTokens + totals.cacheReadTokens;
   const outputTokens = totals.outputTokens;
-  const cacheReadTokens = totals.cacheReadTokens;
-  const cacheCreationTokens = totals.cacheCreationTokens;
-  const reasoningTokens = totals.reasoningTokens;
-  const used = inputTokens + outputTokens + cacheReadTokens + cacheCreationTokens + reasoningTokens;
+  const used = inputTokens
+    + outputTokens
+    + totals.reasoningTokens
+    + totals.cacheReadTokens
+    + totals.cacheWriteTokens;
 
   if (used <= 0) {
     return undefined;
@@ -118,14 +120,50 @@ const buildTokenUsage = (totals: OpenCodeTokenTotals | undefined): AnyRecord | u
 
   return {
     used,
-    total: used,
-    inputTokens,
+    inputTokens: displayInputTokens,
     outputTokens,
-    cacheReadTokens,
-    cacheCreationTokens,
+    breakdown: {
+      input: displayInputTokens,
+      output: outputTokens,
+    },
   };
 };
 
+const readOpenCodeSessionColumnTokenUsage = (
+  db: Database.Database,
+  sessionId: string,
+): AnyRecord | undefined => {
+  const columns = db.prepare('PRAGMA table_info(session)').all() as { name: string }[];
+  const columnNames = new Set(columns.map((column) => column.name));
+  const requiredColumns = ['tokens_input', 'tokens_output', 'tokens_reasoning', 'tokens_cache_read', 'tokens_cache_write'];
+  if (!requiredColumns.every((column) => columnNames.has(column))) {
+    return undefined;
+  }
+
+  const row = db.prepare(`
+    SELECT
+      tokens_input AS inputTokens,
+      tokens_output AS outputTokens,
+      tokens_reasoning AS reasoningTokens,
+      tokens_cache_read AS cacheReadTokens,
+      tokens_cache_write AS cacheWriteTokens
+    FROM session
+    WHERE id = ?
+  `).get(sessionId) as OpenCodeTokenTotals | undefined;
+
+  if (!row) {
+    return undefined;
+  }
+
+  return buildTokenUsage({
+    inputTokens: Number(row.inputTokens ?? 0),
+    outputTokens: Number(row.outputTokens ?? 0),
+    reasoningTokens: Number(row.reasoningTokens ?? 0),
+    cacheReadTokens: Number(row.cacheReadTokens ?? 0),
+    cacheWriteTokens: Number(row.cacheWriteTokens ?? 0),
+  });
+};
+
 /**
  * OpenCode stores per-message token counts on assistant `message.data` objects
  * (see MessageV2.Assistant). Older DBs also had session-level counters; this
@@ -135,13 +173,18 @@ const aggregateOpenCodeSessionTokenUsage = (
   db: Database.Database,
   sessionId: string,
 ): AnyRecord | undefined => {
+  const sessionColumnUsage = readOpenCodeSessionColumnTokenUsage(db, sessionId);
+  if (sessionColumnUsage) {
+    return sessionColumnUsage;
+  }
+
   const rows = db.prepare('SELECT data FROM message WHERE session_id = ?').all(sessionId) as { data: string }[];
 
   let inputTokens = 0;
   let outputTokens = 0;
-  let cacheReadTokens = 0;
-  let cacheCreationTokens = 0;
   let reasoningTokens = 0;
+  let cacheReadTokens = 0;
+  let cacheWriteTokens = 0;
 
   for (const row of rows) {
     const info = readJsonRecord(row.data);
@@ -159,15 +202,15 @@ const aggregateOpenCodeSessionTokenUsage = (
     reasoningTokens += Number(tokens.reasoning ?? 0);
     const cache = readObjectRecord(tokens.cache);
     cacheReadTokens += Number(cache?.read ?? 0);
-    cacheCreationTokens += Number(cache?.write ?? 0);
+    cacheWriteTokens += Number(cache?.write ?? 0);
   }
 
   return buildTokenUsage({
     inputTokens,
     outputTokens,
-    cacheReadTokens,
-    cacheCreationTokens,
     reasoningTokens,
+    cacheReadTokens,
+    cacheWriteTokens,
   });
 };