Get returns the cached value and true if present and unexpired.

Reload config and model modules to pick up env var changes.

Reload config and model modules to pick up env var changes.

Reload config and model modules to pick up env var changes.

Reload config and model modules to pick up env var changes.

List returns the items of a Kind+Group, optionally scoped to a namespace.

AddTool registers a handler wrapped with metrics + rate limiting.

Instantiate and return the configured LLM model. Reads LLM_PROVIDER from environment (via Config). Raises ValueError for unknown provide

Instantiate and return the configured LLM model. Reads LLM_PROVIDER from environment (via Config). Raises ValueError for unknown provide

Instantiate and return the configured LLM model. Reads LLM_PROVIDER from environment (via Config). Raises ValueError for unknown provide

Instantiate and return the configured LLM model. Reads LLM_PROVIDER from environment (via Config). Raises ValueError for unknown provide

Fetch logs from this source. Args: target: log group / pod name / index — interpretation is source-specific

One-line representation for prompt injection.

Allow returns true when the caller is permitted to proceed.

LogK8sOperation logs Kubernetes operations

LogK8sOperation logs Kubernetes operations

ServeStdio runs stdio transport.

Render entries as a clean text block for the agent.

Return the N most recent incidents.

Return the N most recent incidents.

Return the N most recent incidents.

Get resolves a logical cluster name to a Cluster handle.

Extract text content from various response formats. LLM responses can come in different formats: - Plain strings - Objects with

Extract text content from various response formats. LLM responses can come in different formats: - Plain strings - Objects with

Extract text content from various response formats. LLM responses can come in different formats: - Plain strings - Objects with

Extract text content from various response formats. LLM responses can come in different formats: - Plain strings - Objects with

Extract text content from various response formats. LLM responses can come in different formats: - Plain strings - Objects with

Fetch logs from this source. Args: target: log group / pod name / index — interpretation is source-specific

Render entries as a clean text block for the agent.

Format a list of incidents into a text block suitable for injection into the system prompt.

Format a list of incidents into a text block suitable for injection into the system prompt.

Format a list of incidents into a text block suitable for injection into the system prompt.

Simple keyword search across incident fields. Returns incidents whose summary, root_cause, or affected_systems contain the qu

Simple keyword search across incident fields. Returns incidents whose summary, root_cause, or affected_systems contain the qu

validateConfirmation ensures dangerous operations require explicit confirmation

validateConfirmation ensures dangerous operations require explicit confirmation

Default returns the first cluster declared in the registry.

ExecuteTool executes the specified tool with the provided input

FetchAll walks every page up to maxItems. Pass maxItems<=0 for unbounded.

LogMCPResponse logs MCP responses with timing

LogMCPResponse logs MCP responses with timing

Names returns the registered cluster names in registry order.

New constructs a Limiter that refills at `rps` requests/second with `burst` tokens. Entries idle for longer than `ttl` are garbage-collected on the ne

Set inserts or refreshes a value, evicting the LRU entry when full.

Build a structured Slack message payload

Build a structured Slack message payload

Build a structured Slack message payload

Build a structured Slack message payload

Delete the session file.

Delete the session file.

Delete the session file.

Greedy time-window clustering. Walk entries in chronological order. Start a new cluster every time the gap from the previous

Extract text content from various response formats. LLM responses can come in different formats: - Plain strings - Objects with

Extract text content from various response formats.

Extract text content from various response formats. LLM responses can come in different formats: - Plain strings - Objects with

Fetch logs from every configured source in parallel. Args: query: Optional substring/query passed to every so

Helper function to format duration in a human-readable way

Helper function to format duration in a human-readable way

Helper function to format duration in a human-readable way

Get the LLM instance

Get the LLM instance

Get the LLM instance

Get the LLM instance