holmesgpt 0.13.1__tar.gz → 0.13.3a0__tar.gz

{holmesgpt-0.13.1 → holmesgpt-0.13.3a0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: holmesgpt
-Version: 0.13.1
+Version: 0.13.3a0
 Summary: 
 Author: Natan Yellin
 Author-email: natan@robusta.dev
@@ -35,6 +35,7 @@ Requires-Dist: mcp (==v1.12.2)
 Requires-Dist: openai (>=1.6.1,<1.100.0)
 Requires-Dist: opensearch-py (>=2.8.0,<3.0.0)
 Requires-Dist: postgrest (==0.16.8)
+Requires-Dist: prometrix (==0.2.3)
 Requires-Dist: prompt-toolkit (>=3.0.51,<4.0.0)
 Requires-Dist: protobuf (>=6.31.1)
 Requires-Dist: pydantic (>=2.7,<3.0)
@@ -50,7 +51,7 @@ Requires-Dist: requests (>=2.32.4,<3.0.0)
 Requires-Dist: requests-aws4auth (>=1.3.1,<2.0.0)
 Requires-Dist: rich (>=13.7.1,<14.0.0)
 Requires-Dist: sentry-sdk[fastapi] (>=2.20.0,<3.0.0)
-Requires-Dist: setuptools (==70.0.0)
+Requires-Dist: setuptools (>=80.9.0,<81.0.0)
 Requires-Dist: slack-bolt (>=1.18.1,<2.0.0)
 Requires-Dist: starlette (==0.47.2)
 Requires-Dist: strenum (>=0.4.15,<0.5.0)

{holmesgpt-0.13.1 → holmesgpt-0.13.3a0}/holmes/__init__.py

@@ -1,5 +1,5 @@
 # This is patched by github actions during release
-__version__ = "0.13.1"
+__version__ = "0.13.3-alpha"
 
 # Re-export version functions from version module for backward compatibility
 from .version import (

{holmesgpt-0.13.1 → holmesgpt-0.13.3a0}/holmes/common/env_vars.py

@@ -67,3 +67,10 @@ MAX_OUTPUT_TOKEN_RESERVATION = int(
 
 # When using the bash tool, setting BASH_TOOL_UNSAFE_ALLOW_ALL will skip any command validation and run any command requested by the LLM
 BASH_TOOL_UNSAFE_ALLOW_ALL = load_bool("BASH_TOOL_UNSAFE_ALLOW_ALL", False)
+
+LOG_LLM_USAGE_RESPONSE = load_bool("LOG_LLM_USAGE_RESPONSE", False)
+
+# For CLI only, enable user approval for potentially sensitive commands that would otherwise be rejected
+ENABLE_CLI_TOOL_APPROVAL = load_bool("ENABLE_CLI_TOOL_APPROVAL", True)
+
+MAX_GRAPH_POINTS = float(os.environ.get("MAX_GRAPH_POINTS", 300))

{holmesgpt-0.13.1 → holmesgpt-0.13.3a0}/holmes/config.py

@@ -527,7 +527,9 @@ class Config(RobustaBaseConfig):
                 if model_key
                 else next(iter(self._model_list.values())).copy()
             )
-            if model_params.get("is_robusta_model") and self.api_key:
+            is_robusta_model = model_params.pop("is_robusta_model", False)
+            if is_robusta_model and self.api_key:
+                # we set here the api_key since it is being refresh when exprided and not as part of the model loading.
                 api_key = self.api_key.get_secret_value()
             else:
                 api_key = model_params.pop("api_key", api_key)

{holmesgpt-0.13.1 → holmesgpt-0.13.3a0}/holmes/core/conversations.py

@@ -133,7 +133,6 @@ def build_issue_chat_messages(
                     "issue": issue_chat_request.issue_type,
                     "toolsets": ai.tool_executor.toolsets,
                     "cluster_name": config.cluster_name,
-                    "investigation_id": ai.investigation_id,
                 },
             )
             messages = [
@@ -154,7 +153,6 @@ def build_issue_chat_messages(
             "issue": issue_chat_request.issue_type,
             "toolsets": ai.tool_executor.toolsets,
             "cluster_name": config.cluster_name,
-            "investigation_id": ai.investigation_id,
         }
         system_prompt_without_tools = load_and_render_prompt(
             template_path, template_context_without_tools
@@ -188,7 +186,6 @@ def build_issue_chat_messages(
             "issue": issue_chat_request.issue_type,
             "toolsets": ai.tool_executor.toolsets,
             "cluster_name": config.cluster_name,
-            "investigation_id": ai.investigation_id,
         }
         system_prompt_with_truncated_tools = load_and_render_prompt(
             template_path, truncated_template_context
@@ -230,7 +227,6 @@ def build_issue_chat_messages(
         "issue": issue_chat_request.issue_type,
         "toolsets": ai.tool_executor.toolsets,
         "cluster_name": config.cluster_name,
-        "investigation_id": ai.investigation_id,
     }
     system_prompt_without_tools = load_and_render_prompt(
         template_path, template_context_without_tools
@@ -254,7 +250,6 @@ def build_issue_chat_messages(
         "issue": issue_chat_request.issue_type,
         "toolsets": ai.tool_executor.toolsets,
         "cluster_name": config.cluster_name,
-        "investigation_id": ai.investigation_id,
     }
     system_prompt_with_truncated_tools = load_and_render_prompt(
         template_path, template_context
@@ -279,7 +274,6 @@ def add_or_update_system_prompt(
     context = {
         "toolsets": ai.tool_executor.toolsets,
         "cluster_name": config.cluster_name,
-        "investigation_id": ai.investigation_id,
     }
 
     system_prompt = load_and_render_prompt(template_path, context)
@@ -471,7 +465,6 @@ def build_workload_health_chat_messages(
                     "resource": resource,
                     "toolsets": ai.tool_executor.toolsets,
                     "cluster_name": config.cluster_name,
-                    "investigation_id": ai.investigation_id,
                 },
             )
             messages = [
@@ -492,7 +485,6 @@ def build_workload_health_chat_messages(
             "resource": resource,
             "toolsets": ai.tool_executor.toolsets,
             "cluster_name": config.cluster_name,
-            "investigation_id": ai.investigation_id,
         }
         system_prompt_without_tools = load_and_render_prompt(
             template_path, template_context_without_tools
@@ -526,7 +518,6 @@ def build_workload_health_chat_messages(
             "resource": resource,
             "toolsets": ai.tool_executor.toolsets,
             "cluster_name": config.cluster_name,
-            "investigation_id": ai.investigation_id,
         }
         system_prompt_with_truncated_tools = load_and_render_prompt(
             template_path, truncated_template_context
@@ -568,7 +559,6 @@ def build_workload_health_chat_messages(
         "resource": resource,
         "toolsets": ai.tool_executor.toolsets,
         "cluster_name": config.cluster_name,
-        "investigation_id": ai.investigation_id,
     }
     system_prompt_without_tools = load_and_render_prompt(
         template_path, template_context_without_tools
@@ -592,7 +582,6 @@ def build_workload_health_chat_messages(
         "resource": resource,
         "toolsets": ai.tool_executor.toolsets,
         "cluster_name": config.cluster_name,
-        "investigation_id": ai.investigation_id,
     }
     system_prompt_with_truncated_tools = load_and_render_prompt(
         template_path, template_context

{holmesgpt-0.13.1 → holmesgpt-0.13.3a0}/holmes/core/investigation.py

@@ -9,7 +9,6 @@ from holmes.core.models import InvestigateRequest, InvestigationResult
 from holmes.core.supabase_dal import SupabaseDal
 from holmes.core.tracing import DummySpan, SpanType
 from holmes.utils.global_instructions import add_global_instructions_to_user_prompt
-from holmes.core.todo_manager import get_todo_manager
 
 from holmes.core.investigation_structured_output import (
     DEFAULT_SECTIONS,
@@ -133,9 +132,6 @@ def get_investigation_context(
     else:
         logging.info("Structured output is disabled for this request")
 
-    todo_manager = get_todo_manager()
-    todo_context = todo_manager.format_tasks_for_prompt(ai.investigation_id)
-
     system_prompt = load_and_render_prompt(
         investigate_request.prompt_template,
         {
@@ -144,8 +140,6 @@ def get_investigation_context(
             "structured_output": request_structured_output_from_llm,
             "toolsets": ai.tool_executor.toolsets,
             "cluster_name": config.cluster_name,
-            "todo_list": todo_context,
-            "investigation_id": ai.investigation_id,
         },
     )
 

{holmesgpt-0.13.1 → holmesgpt-0.13.3a0}/holmes/core/llm.py

@@ -229,9 +229,11 @@ class DefaultLLM(LLM):
             ]  # can be removed after next litelm version
 
         self.args.setdefault("temperature", temperature)
+
+        self._add_cache_control_to_last_message(messages)
+
         # Get the litellm module to use (wrapped or unwrapped)
         litellm_to_use = self.tracer.wrap_llm(litellm) if self.tracer else litellm
-
         result = litellm_to_use.completion(
             model=self.model,
             api_key=self.api_key,
@@ -266,3 +268,60 @@ class DefaultLLM(LLM):
                 f"Couldn't find model's name {model_name} in litellm's model list, fallback to 4096 tokens for max_output_tokens"
             )
             return 4096
+
+    def _add_cache_control_to_last_message(
+        self, messages: List[Dict[str, Any]]
+    ) -> None:
+        """
+        Add cache_control to the last non-user message for Anthropic prompt caching.
+        Removes any existing cache_control from previous messages to avoid accumulation.
+        """
+        # First, remove any existing cache_control from all messages
+        for msg in messages:
+            content = msg.get("content")
+            if isinstance(content, list):
+                for block in content:
+                    if isinstance(block, dict) and "cache_control" in block:
+                        del block["cache_control"]
+                        logging.debug(
+                            f"Removed existing cache_control from {msg.get('role')} message"
+                        )
+
+        # Find the last non-user message to add cache_control to.
+        # Adding cache_control to user message requires changing its structure, so we avoid it
+        # This avoids breaking parse_messages_tags which only processes user messages
+        target_msg = None
+        for msg in reversed(messages):
+            if msg.get("role") != "user":
+                target_msg = msg
+                break
+
+        if not target_msg:
+            logging.debug("No non-user message found for cache_control")
+            return
+
+        content = target_msg.get("content")
+
+        if content is None:
+            return
+
+        if isinstance(content, str):
+            # Convert string to structured format with cache_control
+            target_msg["content"] = [
+                {
+                    "type": "text",
+                    "text": content,
+                    "cache_control": {"type": "ephemeral"},
+                }
+            ]
+            logging.debug(
+                f"Added cache_control to {target_msg.get('role')} message (converted from string)"
+            )
+        elif isinstance(content, list) and content:
+            # Add cache_control to the last content block
+            last_block = content[-1]
+            if isinstance(last_block, dict) and "type" in last_block:
+                last_block["cache_control"] = {"type": "ephemeral"}
+                logging.debug(
+                    f"Added cache_control to {target_msg.get('role')} message (structured content)"
+                )

{holmesgpt-0.13.1 → holmesgpt-0.13.3a0}/holmes/core/prompt.py

@@ -40,7 +40,6 @@ def build_initial_ask_messages(
     initial_user_prompt: str,
     file_paths: Optional[List[Path]],
     tool_executor: Any,  # ToolExecutor type
-    investigation_id: str,
     runbooks: Union[RunbookCatalog, Dict, None] = None,
     system_prompt_additions: Optional[str] = None,
 ) -> List[Dict]:
@@ -60,7 +59,6 @@ def build_initial_ask_messages(
         "toolsets": tool_executor.toolsets,
         "runbooks": runbooks or {},
         "system_prompt_additions": system_prompt_additions or "",
-        "investigation_id": investigation_id,
     }
     system_prompt_rendered = load_and_render_prompt(
         system_prompt_template, template_context

{holmesgpt-0.13.1 → holmesgpt-0.13.3a0}/holmes/core/supabase_dal.py

@@ -131,7 +131,7 @@ class SupabaseDal:
                         raise Exception(
                             "No robusta token provided to Holmes.\n"
                             "Please set a valid Robusta UI token.\n "
-                            "See https://docs.robusta.dev/master/configuration/ai-analysis.html#choosing-and-configuring-an-ai-provider for instructions."
+                            "See https://holmesgpt.dev/ai-providers/ for instructions."
                         )
                     env_replacement_token = get_env_replacement(token)
                     if env_replacement_token:
@@ -143,7 +143,7 @@ class SupabaseDal:
                             "Ensure your Helm chart or environment variables are set correctly.\n "
                             "If you store the token in a secret, you must also pass "
                             "the environment variable ROBUSTA_UI_TOKEN to Holmes.\n "
-                            "See https://docs.robusta.dev/master/configuration/ai-analysis.html#configuring-holmesgpt-access-to-saas-data for instructions."
+                            "See https://holmesgpt.dev/data-sources/builtin-toolsets/robusta/ for instructions."
                         )
                     try:
                         decoded = base64.b64decode(token)

holmesgpt-0.13.3a0/holmes/core/todo_tasks_formatter.py

@@ -0,0 +1,51 @@
+from typing import List
+
+from holmes.plugins.toolsets.investigator.model import Task, TaskStatus
+
+
+def format_tasks(tasks: List[Task]) -> str:
+    """
+    Format tasks for tool response
+    Returns empty string if no tasks exist.
+    """
+    if not tasks:
+        return ""
+
+    status_order = {
+        TaskStatus.PENDING: 0,
+        TaskStatus.IN_PROGRESS: 1,
+        TaskStatus.COMPLETED: 2,
+    }
+
+    sorted_tasks = sorted(
+        tasks,
+        key=lambda t: (status_order.get(t.status, 3),),
+    )
+
+    lines = ["# CURRENT INVESTIGATION TASKS"]
+    lines.append("")
+
+    pending_count = sum(1 for t in tasks if t.status == TaskStatus.PENDING)
+    progress_count = sum(1 for t in tasks if t.status == TaskStatus.IN_PROGRESS)
+    completed_count = sum(1 for t in tasks if t.status == TaskStatus.COMPLETED)
+
+    lines.append(
+        f"**Task Status**: {completed_count} completed, {progress_count} in progress, {pending_count} pending"
+    )
+    lines.append("")
+
+    for task in sorted_tasks:
+        status_indicator = {
+            TaskStatus.PENDING: "[ ]",
+            TaskStatus.IN_PROGRESS: "[~]",
+            TaskStatus.COMPLETED: "[✓]",
+        }.get(task.status, "[?]")
+
+        lines.append(f"{status_indicator} [{task.id}] {task.content}")
+
+    lines.append("")
+    lines.append(
+        "**Instructions**: Use TodoWrite tool to update task status as you work. Mark tasks as 'in_progress' when starting, 'completed' when finished."
+    )
+
+    return "\n".join(lines)