OpenAI-compatible HTTP API: stdlib-only gateway over SERVE with KV prefix reuse across stateless requests (#21)

* Add OpenAI-compatible HTTP API

* Support browser API clients

* Handle missing KV cache during rewind

c/coli

@@ -4,6 +4,7 @@ colibrì — piccolo motore, modello immenso.
 CLI per far girare GLM-5.2 (744B) in locale, su CPU, in ~15-26 GB di RAM.
 
   coli chat                 chat interattiva (carica il modello UNA volta)
+  coli serve                API HTTP compatibile OpenAI (motore persistente)
   coli run "prompt"         generazione singola
   coli info                 stato: modello, RAM, disco, config
   coli bench [task...]      benchmark di qualità (MMLU/HellaSwag/...)
@@ -375,6 +376,12 @@ def cmd_chat(a):
         except Exception: pass
     print(f"  {C.teal}ciao{C.r} {C.dim}— il colibrì torna al nido{C.r} 🐦\n")
 
+def cmd_serve(a):
+    need_model(a.model)
+    from openai_server import serve
+    serve(a.model, a.host, a.port, a.model_id, a.api_key,
+          a.cap, a.ngen, GLM, env_for(a), a.cors_origin)
+
 def cmd_bench(a):
     need_model(a.model)
     banner("bench")
@@ -427,13 +434,18 @@ def main():
     sub.add_parser("build", parents=[common]); sub.add_parser("info", parents=[common])
     pr=sub.add_parser("run", parents=[common]);  pr.add_argument("prompt", nargs="*")
     sub.add_parser("chat", parents=[common])
+    ps=sub.add_parser("serve", parents=[common])
+    ps.add_argument("--host",default="127.0.0.1"); ps.add_argument("--port",type=int,default=8000)
+    ps.add_argument("--model-id",default=os.environ.get("COLI_MODEL_ID","glm-5.2-colibri"))
+    ps.add_argument("--api-key",default=os.environ.get("COLI_API_KEY"))
+    ps.add_argument("--cors-origin",action="append",default=None)
     pb=sub.add_parser("bench", parents=[common]); pb.add_argument("tasks", nargs="*")
     pb.add_argument("--limit",type=int,default=40); pb.add_argument("--data",default=os.path.join(HERE,"bench"))
     pc=sub.add_parser("convert", parents=[common]); pc.add_argument("--repo",default="zai-org/GLM-5.2-FP8")
     pc.add_argument("--ebits",type=int,default=4); pc.add_argument("--io-bits",type=int,default=8); pc.add_argument("--xbits",type=int,default=0)
     pc.add_argument("--no-mtp",action="store_true",help="salta la testa MTP (niente draft speculativi)")
     a=ap.parse_args()
-    {"build":cmd_build,"info":cmd_info,"run":cmd_run,"chat":cmd_chat,"bench":cmd_bench,
+    {"build":cmd_build,"info":cmd_info,"run":cmd_run,"chat":cmd_chat,"serve":cmd_serve,"bench":cmd_bench,
      "convert":cmd_convert}.get(a.cmd, lambda _:(banner(),print(__doc__)))(a)
 
 if __name__=="__main__":