opencode-llmstack 0.8.0__tar.gz → 0.9.1__tar.gz

{opencode_llmstack-0.8.0 → opencode_llmstack-0.9.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: opencode-llmstack
-Version: 0.8.0
+Version: 0.9.1
 Summary: Multi-tier local LLM stack: llama-swap + FastAPI auto-router + opencode wiring.
 Author: llmstack
 License: MIT

{opencode_llmstack-0.8.0 → opencode_llmstack-0.9.1}/llmstack/app.py

@@ -90,16 +90,13 @@ Routing decision tree (first match wins):
      ("reasonable context still being built")         -> code-ultra
                                                          (else code-smart)
   5. Estimated input tokens <= MID_FIDELITY_CEILING   -> code-smart
-  6. Otherwise (long context, top-tier becomes
-     expensive/slow, fast tier's 128k window is the
-     best fit and it's free)                          -> code-fast
+   6. Otherwise (long context, top-tier becomes
+      expensive/slow, fast tier's 128k window is the
+      best fit and it's free)                          -> code-fast
                                                          (floored at
                                                           code-smart when
-                                                          ``tools[]`` is set
-                                                          or n_turns >=
-                                                          MULTI_TURN_THRESHOLD,
-                                                          since 3B models
-                                                          tool-call unreliably)
+                                                          n_turns >=
+                                                          MULTI_TURN_THRESHOLD)
 
 The auto router's effective max context window is
 ``[code-fast].ctx_size`` -- fast is the bottom of the step-down
@@ -173,7 +170,7 @@ MID_FIDELITY_CEILING = int(os.getenv("ROUTER_MID_FIDELITY_CEILING", "32000"))
 # Floor the long-context rung at code-smart whenever a tool-call
 # protocol is in play -- 3B models tool-call unreliably regardless of
 # how big their context window is.
-MULTI_TURN_THRESHOLD = int(os.getenv("ROUTER_MULTI_TURN", "6"))
+MULTI_TURN_THRESHOLD = int(os.getenv("ROUTER_MULTI_TURN", "10"))
 AUTO_ALIASES = {"auto", "", None}
 
 UNCENSORED_TRIGGERS = re.compile(
@@ -356,8 +353,7 @@ def classify(body: dict[str, Any]) -> tuple[str, str]:
                     ULTRA_MODEL, AGENT_MODEL)
         return AGENT_MODEL, f"ultra-trigger->agent ({ULTRA_MODEL} unavailable)"
 
-    has_tools = bool(body.get("tools"))
-    n_turns = len(messages) if messages else 0
+    n_turns = sum(1 for m in (messages or []) if m.get("role") == "user")
     has_code_signal = (
         _matches(CODE_BLOCK, messages, prompt)
         or _matches(AGENT_SIGNALS, messages, prompt)
@@ -368,14 +364,14 @@ def classify(body: dict[str, Any]) -> tuple[str, str]:
     # Plan track is orthogonal to the code fidelity ladder: ``plan`` is a
     # chat-tuned model meant for design / "should we" discussions. Only
     # take it when nothing about the request says "I'm about to write
-    # code" (no triple-backticks, no agent verbs, no tool calls). And
-    # only if the input fits in the planner's ctx_size -- past that we'd
-    # be sending a request the planner can't hold, so we fall through
-    # to the coding ladder, which has tiers (smart, fast) explicitly
-    # sized for larger contexts.
+    # code" (no triple-backticks, no agent verbs). Tools are stripped
+    # from the request body before dispatch (see ``_handle_completion``),
+    # so their presence here does not block plan routing.
+    # Only route to plan if the input fits in the planner's ctx_size --
+    # past that we fall through to the coding ladder which has tiers
+    # (smart, fast) explicitly sized for larger contexts.
     if (
-        not has_tools
-        and not has_code_signal
+        not has_code_signal
         and _matches(PLAN_SIGNALS, messages, prompt)
     ):
         plan_tier = TIER_BY_ALIAS.get(PLAN_MODEL)
@@ -401,12 +397,12 @@ def classify(body: dict[str, Any]) -> tuple[str, str]:
     if est <= MID_FIDELITY_CEILING:
         return AGENT_MODEL, f"mid-fidelity tokens~{est}<={MID_FIDELITY_CEILING}"
 
-    # Rung 3: long context -- step down to fast (128k YaRN, free,
-    # always-resident). Floor at smart when tools/agent loop is in
-    # play; the 3B coder doesn't tool-call reliably.
-    if has_tools or n_turns >= MULTI_TURN_THRESHOLD:
-        why = "tools" if has_tools else f"turns={n_turns}"
-        return AGENT_MODEL, f"long-context tokens~{est}>{MID_FIDELITY_CEILING} ({why} floor)"
+    # Rung 3: long context -- step down to fast. Floor at smart only
+    # when the multi-turn threshold is hit; tools alone no longer
+    # prevent the step-down (plan tiers strip tools before dispatch,
+    # and code-fast is a hosted model that tool-calls reliably).
+    if n_turns >= MULTI_TURN_THRESHOLD:
+        return AGENT_MODEL, f"long-context tokens~{est}>{MID_FIDELITY_CEILING} (user-turns={n_turns}>={MULTI_TURN_THRESHOLD} floor)"
     return FAST_MODEL, f"long-context tokens~{est}>{MID_FIDELITY_CEILING}"
 
 
@@ -626,6 +622,11 @@ async def _handle_completion(req: Request, path: str) -> Response:
         mutated = True
 
     chosen_name = body.get("model")
+    if chosen_name in {PLAN_MODEL, UNCENSORED_MODEL} and body.get("tools"):
+        log.info("plan tier %s: stripping tools from request", chosen_name)
+        body.pop("tools")
+        body.pop("tool_choice", None)
+        mutated = True
     tier = _resolve_tier(chosen_name)
     if tier is not None and _inject_sampler(body, tier):
         mutated = True

{opencode_llmstack-0.8.0 → opencode_llmstack-0.9.1}/llmstack/models.ini

@@ -300,7 +300,7 @@ description  = Mistral-Small 3.2 24B Heretic - no-filter planning
 ;
 high_fidelity_ceiling = 12000    ; tokens; below this, top-tier model is still cheap+fast (and ultra ctx_size = 2 * this)
 mid_fidelity_ceiling  = 32000    ; tokens; smart's sweet spot up to here, then step down to fast (smart ctx_size = 2 * this)
-multi_turn            = 6        ; turn count that floors the long-context rung at code-smart
+multi_turn            = 10       ; turn count that floors the long-context rung at code-smart
 agent_signal_words    = implement, fix bug, write a function, refactor, edit, patch, debug, run tests, build it
 plan_signal_words     = design, architect, approach, trade-off, should we, how would you, explain why, think through, compare options, brainstorm, root cause
 uncensored_triggers   = [nofilter], [uncensored], [heretic], "uncensored:", "nofilter:" (line start)

{opencode_llmstack-0.8.0 → opencode_llmstack-0.9.1}/opencode_llmstack.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: opencode-llmstack
-Version: 0.8.0
+Version: 0.9.1
 Summary: Multi-tier local LLM stack: llama-swap + FastAPI auto-router + opencode wiring.
 Author: llmstack
 License: MIT

{opencode_llmstack-0.8.0 → opencode_llmstack-0.9.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "opencode-llmstack"
-version = "0.8.0"
+version = "0.9.1"
 description = "Multi-tier local LLM stack: llama-swap + FastAPI auto-router + opencode wiring."
 readme = "README.md"
 requires-python = ">=3.11"