fix: allow memory flush model override

docs/concepts/compaction.md

@@ -132,7 +132,23 @@ By default, compaction runs silently. Set `notifyUser` to show brief status mess
 
 ### Memory flush
 
-Before compaction, OpenClaw can run a **silent memory flush** turn to store durable notes to disk. See [Memory](/concepts/memory) for details and config.
+Before compaction, OpenClaw can run a **silent memory flush** turn to store durable notes to disk. Set `agents.defaults.compaction.memoryFlush.model` when this housekeeping turn should use a local model instead of the active conversation model:
+
+```json
+{
+  "agents": {
+    "defaults": {
+      "compaction": {
+        "memoryFlush": {
+          "model": "ollama/qwen3:8b"
+        }
+      }
+    }
+  }
+}
+```
+
+The memory-flush model override is exact and does not inherit the active session fallback chain. See [Memory](/concepts/memory) for details and config.
 
 ## Pluggable compaction providers
 

docs/concepts/memory.md

@@ -110,6 +110,26 @@ Before [compaction](/concepts/compaction) summarizes your conversation, OpenClaw
 runs a silent turn that reminds the agent to save important context to memory
 files. This is on by default — you do not need to configure anything.
 
+To keep that housekeeping turn on a local model, set an exact memory-flush model
+override:
+
+```json
+{
+  "agents": {
+    "defaults": {
+      "compaction": {
+        "memoryFlush": {
+          "model": "ollama/qwen3:8b"
+        }
+      }
+    }
+  }
+}
+```
+
+The override applies only to the memory-flush turn and does not inherit the
+active session fallback chain.
+
 <Tip>
 The memory flush prevents context loss during compaction. If your agent has
 important facts in the conversation that are not yet written to a file, they

docs/gateway/config-agents.md

@@ -559,6 +559,7 @@ Periodic heartbeat runs.
         notifyUser: true, // send brief notices when compaction starts and completes (default: false)
         memoryFlush: {
           enabled: true,
+          model: "ollama/qwen3:8b", // optional memory-flush-only model override
           softThresholdTokens: 6000,
           systemPrompt: "Session nearing compaction. Store durable memories now.",
           prompt: "Write any lasting notes to memory/YYYY-MM-DD.md; reply with the exact silent token NO_REPLY if nothing to store.",
@@ -580,7 +581,7 @@ Periodic heartbeat runs.
 - `model`: optional `provider/model-id` override for compaction summarization only. Use this when the main session should keep one model but compaction summaries should run on another; when unset, compaction uses the session's primary model.
 - `maxActiveTranscriptBytes`: optional byte threshold (`number` or strings like `"20mb"`) that triggers normal local compaction before a run when the active JSONL grows past the threshold. Requires `truncateAfterCompaction` so successful compaction can rotate to a smaller successor transcript. Disabled when unset or `0`.
 - `notifyUser`: when `true`, sends brief notices to the user when compaction starts and when it completes (for example, "Compacting context..." and "Compaction complete"). Disabled by default to keep compaction silent.
-- `memoryFlush`: silent agentic turn before auto-compaction to store durable memories. Skipped when workspace is read-only.
+- `memoryFlush`: silent agentic turn before auto-compaction to store durable memories. Set `model` to an exact provider/model such as `ollama/qwen3:8b` when this housekeeping turn should stay on a local model; the override does not inherit the active session fallback chain. Skipped when workspace is read-only.
 
 ### `agents.defaults.contextPruning`
 

docs/plugins/sdk-overview.md

@@ -273,6 +273,9 @@ AI CLI backend such as `codex-cli`.
   memory plugin's private layout.
 - `registerMemoryPromptSection`, `registerMemoryFlushPlan`, and
   `registerMemoryRuntime` are legacy-compatible exclusive memory-plugin APIs.
+- `MemoryFlushPlan.model` can pin the flush turn to an exact `provider/model`
+  reference, such as `ollama/qwen3:8b`, without inheriting the active fallback
+  chain.
 - `registerMemoryEmbeddingProvider` lets the active memory plugin register one
   or more embedding adapter ids (for example `openai`, `gemini`, or a custom
   plugin-defined id).

docs/reference/session-management-compaction.md

@@ -381,6 +381,7 @@ OpenClaw uses the **pre-threshold flush** approach:
 Config (`agents.defaults.compaction.memoryFlush`):
 
 - `enabled` (default: `true`)
+- `model` (optional exact provider/model override for the flush turn, for example `ollama/qwen3:8b`)
 - `softThresholdTokens` (default: `4000`)
 - `prompt` (user message for the flush turn)
 - `systemPrompt` (extra system prompt appended for the flush turn)
@@ -389,6 +390,9 @@ Notes:
 
 - The default prompt/system prompt include a `NO_REPLY` hint to suppress
   delivery.
+- When `model` is set, the flush turn uses that model without inheriting the
+  active session fallback chain, so local-only housekeeping does not silently
+  fall back to a paid conversation model.
 - The flush runs once per compaction cycle (tracked in `sessions.json`).
 - The flush runs only for embedded Pi sessions (CLI backends skip it).
 - The flush is skipped when the session workspace is read-only (`workspaceAccess: "ro"` or `"none"`).