fix: resolve P1 backward compatibility and routing issues

- Fix ManagedAgent compute provider routing: maintain backward compatibility
  for e2b/modal/etc by routing to LocalManagedAgent with deprecation warning
  instead of ValueError (fixes hard breaking change)
- Fix LocalAgent provider preservation: preserve provider value for LLM routing
  to maintain _resolve_model() prefix behavior (ollama/, gemini/)
- Fix spurious deprecation warnings: auto-detected provider=local no longer
  triggers warnings for users who never passed provider=
- Remove unused os import from hosted_agent.py
- Improve error messages with provider-specific guidance
- Fix test fragility with exact warning counts and identity checks

Addresses all P1 issues identified by Greptile and CodeRabbit reviewers.

Co-authored-by: Mervin Praison <MervinPraison@users.noreply.github.com>

Author: praisonai-triage-agent[bot]

src/praisonai/praisonai/integrations/hosted_agent.py

@@ -28,7 +28,6 @@
     - Tools are co-located with the provider infrastructure
 """
 
-import os
 from typing import Optional, Any
 from .managed_agents import AnthropicManagedAgent, ManagedConfig
 
@@ -64,11 +63,30 @@ def __init__(
         **kwargs,
     ):
         if provider != "anthropic":
+            # Provide differentiated guidance based on provider type
+            _llm_hints = {"openai", "gemini", "ollama", "local"}
+            _compute_hints = {"e2b", "modal", "flyio", "daytona", "docker"}
+            
+            if provider in _llm_hints:
+                hint = (
+                    f"For local agent loops with this LLM, use: "
+                    f"LocalAgent(config=LocalAgentConfig(model='...')) "
+                    f"(e.g. 'gpt-4o-mini', 'gemini/gemini-2.0-flash', 'ollama/llama3')."
+                )
+            elif provider in _compute_hints:
+                hint = (
+                    f"For local execution with cloud compute, use: "
+                    f"LocalAgent(compute='{provider}', config=LocalAgentConfig(...))"
+                )
+            else:
+                hint = (
+                    "Use LocalAgent(config=LocalAgentConfig(model='...')) for local loops, "
+                    "or LocalAgent(compute='e2b'|'modal'|'docker'|...) for cloud-sandboxed tools."
+                )
+            
             raise ValueError(
                 f"Managed runtime for provider '{provider}' is not yet available. "
-                f"Currently supported: 'anthropic'. "
-                f"For local execution with cloud compute, use: "
-                f"LocalAgent(compute='{provider}', config=LocalAgentConfig(...))"
+                f"Currently supported: 'anthropic'. {hint}"
             )
 
         # Pass through to the existing Anthropic implementation

src/praisonai/praisonai/integrations/local_agent.py

@@ -67,6 +67,7 @@ def __init__(
         **kwargs,
     ):
         # Reject the provider= overload pattern to force clean usage
+        provider_for_routing = "local"  # Default provider for model routing
         if 'provider' in kwargs:
             provider_value = kwargs.pop('provider')
             warnings.warn(
@@ -77,6 +78,9 @@ def __init__(
                 DeprecationWarning,
                 stacklevel=2
             )
+            # Preserve the provider value for LLM routing to maintain backward compatibility
+            # This ensures _resolve_model() can still apply proper prefixes (ollama/, gemini/, etc.)
+            provider_for_routing = provider_value
 
-        # Pass compute= as the compute parameter to the underlying implementation
-        super().__init__(compute=compute, config=config, **kwargs)
+        # Pass compute= as the compute parameter and provider for LLM routing
+        super().__init__(compute=compute, config=config, provider=provider_for_routing, **kwargs)

src/praisonai/praisonai/integrations/managed_agents.py

@@ -1097,7 +1097,9 @@ def ManagedAgent(
     Raises:
         ValueError: For compute-provider names that should use LocalAgent(compute=).
     """
-    if provider is None:
+    # Track if provider was auto-detected to avoid spurious deprecation warnings
+    auto_detected = provider is None
+    if auto_detected:
         # Auto-detect
         if os.getenv("ANTHROPIC_API_KEY") or os.getenv("CLAUDE_API_KEY"):
             provider = "anthropic"
@@ -1108,24 +1110,28 @@ def ManagedAgent(
     if provider == "anthropic":
         return AnthropicManagedAgent(provider=provider, **kwargs)
 
-    # Compute provider names should use LocalAgent(compute=)
+    # Compute provider names - maintain backward compatibility by passing to LocalManagedAgent
     elif provider in {"e2b", "modal", "flyio", "daytona", "docker"}:
-        raise ValueError(
-            f"Cloud compute provider '{provider}' belongs on LocalAgent(compute='{provider}', ...). "
-            f"Hosted runtimes for these providers are not yet available. "
-            f"Use: LocalAgent(compute='{provider}', config=LocalAgentConfig(model='gpt-4o-mini'))"
-        )
-    
-    # LLM routing hints (deprecated usage)
-    elif provider in {"openai", "gemini", "ollama", "local"}:
         warnings.warn(
-            f"ManagedAgent(provider='{provider}') is deprecated. "
-            f"Use LocalAgent directly with model= instead: "
-            f"LocalAgent(config=LocalAgentConfig(model='gpt-4o-mini'))",
+            f"ManagedAgent(provider='{provider}') for compute providers is deprecated. "
+            f"Use LocalAgent(compute='{provider}', config=LocalAgentConfig(...)) instead.",
             DeprecationWarning,
             stacklevel=2
         )
         from .managed_local import LocalManagedAgent
+        return LocalManagedAgent(provider="local", compute=provider, **kwargs)
+    
+    # LLM routing hints (deprecated usage) - only warn if explicitly passed by user
+    elif provider in {"openai", "gemini", "ollama", "local"}:
+        if not auto_detected:
+            warnings.warn(
+                f"ManagedAgent(provider='{provider}') is deprecated. "
+                f"Use LocalAgent directly with model= instead: "
+                f"LocalAgent(config=LocalAgentConfig(model='gpt-4o-mini'))",
+                DeprecationWarning,
+                stacklevel=2
+            )
+        from .managed_local import LocalManagedAgent
         return LocalManagedAgent(provider=provider, **kwargs)
 
     # Unknown provider

tests/unit/integrations/test_backend_semantics.py

@@ -68,10 +68,14 @@ def test_local_agent_rejects_provider_overload():
     with warnings.catch_warnings(record=True) as w:
         warnings.simplefilter("always")
         LocalAgent(provider="openai", config=LocalAgentConfig(model="gpt-4o-mini"))
-        assert len(w) == 1
-        assert issubclass(w[0].category, DeprecationWarning)
-        assert "provider=" in str(w[0].message)
-        assert "config.model=" in str(w[0].message)
+        # Filter to only DeprecationWarning containing provider= to avoid false positives
+        dep_warnings = [
+            rec for rec in w
+            if issubclass(rec.category, DeprecationWarning)
+            and "provider=" in str(rec.message)
+            and "config.model=" in str(rec.message)
+        ]
+        assert len(dep_warnings) == 1, f"Expected 1 provider= deprecation warning, got {len(dep_warnings)} from {len(w)} total warnings"
 
 
 def test_managed_agent_deprecation_warnings():
@@ -171,10 +175,10 @@ def test_config_aliases():
     from praisonai.integrations.managed_local import LocalManagedConfig
 
     # HostedAgentConfig should alias ManagedConfig
-    assert HostedAgentConfig == ManagedConfig
+    assert HostedAgentConfig is ManagedConfig
 
     # LocalAgentConfig should alias LocalManagedConfig  
-    assert LocalAgentConfig == LocalManagedConfig
+    assert LocalAgentConfig is LocalManagedConfig
 
 
 def test_unknown_provider_error():