feat(server): provider switching API, WS/REST cancel, summarize endpoints

config.example.yaml

@@ -17,11 +17,23 @@ llm:
   model: ""                     # e.g. MiniMaxAI/MiniMax-M2.7, zai-org/GLM-5.3-Flash
   timeout_sec: 120
   temperature: 0.7
+  # Further providers the app may switch to at runtime (PUT /api/providers/active or per session).
+  # Keys stay on the server and are never sent to the app. Presets also pick up their env key.
+  providers:
+    # dahl: { api_key: "" }               # or env DAHL_API_KEY
+    # gonkarouter: { api_key: "" }        # or env GONKAROUTER_API_KEY
+    # mylocal: { base_url: "http://127.0.0.1:8000/v1", model: "qwen3" }
 
 agent:
   system_prompt: "You are ByteClaw, a concise and helpful self-hosted assistant. Use tools when they help."
   max_iterations: 8
   history_limit: 50
+  # Summarise long sessions with the LLM (originals stay in SQLite). Manual: POST /api/sessions/{id}/summarize
+  summarize:
+    enabled: true
+    max_messages: 40     # unsummarised messages before summarising
+    max_tokens: 12000    # or estimated tokens (chars/4)
+    keep_recent: 10      # newest messages always kept verbatim
 
 tools:
   web_fetch: