refactor(web): always enable static-prefix prompt caching

Remove the SOURCEBOT_CHAT_PROMPT_CACHE_STATIC_PREFIX_ENABLED lever so the static checkpoint is always emitted, and switch the remaining cache env vars to JSDoc comments so their descriptions render on IDE hover.

Author: jsourcebot

packages/shared/src/env.server.ts

@@ -311,15 +311,13 @@ const options = {
 
         SOURCEBOT_CHAT_MAX_STEP_COUNT: numberSchema.default(100),
         SOURCEBOT_CHAT_PROMPT_CACHING_ENABLED: booleanSchema.default('true'),
-        // Phased-rollout lever for the static checkpoint. Set to 'false' to fall
-        // back to the single moving tail marker. Only takes effect when prompt
-        // caching is enabled.
-        SOURCEBOT_CHAT_PROMPT_CACHE_STATIC_PREFIX_ENABLED: booleanSchema.default('true'),
-        // TTL for the static block. The moving tail marker always uses the 5m default.
+        /** TTL for the static block. The moving tail marker always uses the 5m default. */
         SOURCEBOT_CHAT_PROMPT_CACHE_STATIC_TTL: z.enum(['5m', '1h']).default('5m'),
-        // Observability: when enabled, logs a warning on unexpected prompt-cache
-        // breaks (static-prefix signature changes, or zero cache reads on a
-        // continuation step). Does not affect request behavior.
+        /**
+         * Observability: when enabled, logs a warning on unexpected prompt-cache
+         * breaks (static-prefix signature changes, or zero cache reads on a
+         * continuation step). Does not affect request behavior.
+         */
         SOURCEBOT_CHAT_PROMPT_CACHE_BREAK_DETECTION_ENABLED: booleanSchema.default('false'),
         SOURCEBOT_MCP_TOOL_CALL_TIMEOUT_MS: numberSchema.int().positive().max(maxTimerDelayMs).default(60000),
 

packages/web/src/ee/features/chat/agent.test.ts

@@ -27,10 +27,6 @@ vi.mock('@sourcebot/shared', () => ({
         SOURCEBOT_CHAT_MAX_STEP_COUNT: 8,
         SOURCEBOT_CHAT_MODEL_TEMPERATURE: 0,
         SOURCEBOT_TELEMETRY_PII_COLLECTION_ENABLED: 'false',
-        // Enable the static front checkpoint so marker placement is exercised;
-        // whether markers are actually emitted is then controlled by the
-        // PromptCacheStrategy passed into each call.
-        SOURCEBOT_CHAT_PROMPT_CACHE_STATIC_PREFIX_ENABLED: 'true',
         SOURCEBOT_CHAT_PROMPT_CACHE_STATIC_TTL: '5m',
         SOURCEBOT_CHAT_PROMPT_CACHE_BREAK_DETECTION_ENABLED: 'false',
     },

packages/web/src/ee/features/chat/agent.ts

@@ -391,10 +391,6 @@ const createAgentStream = async ({
     const mcpRegistry = buildMcpToolRegistry(mcpToolSetsObj.tools);
     const hasMcpTools = mcpRegistry.length > 0;
 
-    // Phased-rollout lever for the static checkpoint. When disabled, only the
-    // moving tail marker is emitted and behavior collapses to the prior
-    // single-breakpoint scheme.
-    const useStaticPrefix = env.SOURCEBOT_CHAT_PROMPT_CACHE_STATIC_PREFIX_ENABLED === 'true';
     const staticTtl = env.SOURCEBOT_CHAT_PROMPT_CACHE_STATIC_TTL;
 
     const toolRequestActivation = tool({
@@ -455,9 +451,7 @@ const createAgentStream = async ({
     // Caveat: when MCP tools are lazily activated mid-run via prepareStep, the
     // tools section grows and invalidates both breakpoints for that step; the
     // cache re-warms on subsequent steps once the active tool set is stable.
-    const staticMarker = useStaticPrefix
-        ? promptCacheStrategy.cacheControl({ ttl: staticTtl })
-        : undefined;
+    const staticMarker = promptCacheStrategy.cacheControl({ ttl: staticTtl });
     const systemMessages: SystemModelMessage[] = [
         { role: 'system', content: staticPrompt, providerOptions: staticMarker },
     ];