Format pseudocode listing for AI decompilation without address prefixes, but include function addresses in headers.

Format pseudocode listing for realtime analysis.

Gets or creates the netnode for storing AI decompilations.

Get combined session and variable summary.

Validate config keys and values without mutating config.

Get the default criteria for function selection.

Launch full binary indexing on a background thread.

Execute actions through MCP or custom implementation.

Parse LLM tool calls into :class:`FunctionEntry` objects.

Populate callees for a function item.

Common function for running realtime analysis with different prompts.

Return sanitized config and the list of detected issues.

Show or update the AI decompilation tab for a function.

Split *func_list* into batches respecting config and token budget (greedily).

Create a new ctx/verbose path pair for this analysis run.

Simply return clean pseudocode with address prefixes for lines that have addresses, and 'cannotComment' for lines without addresses.

Simply return clean pseudocode with address prefixes for lines that have addresses, and 'cannotComment' for lines without addresses.

Clean pseudocode for realtime analysis.

Summarize a batch of conversation blocks using LLM. Args: blocks: List of ConversationBlock objects to summarize

Summarize a list of conversation turns. Args: turns: Turns to summarize llm_client: Optional LLM cli

Writes the analysis history to the netnode.

Set the current function address and update display.

Show the form.

Show the form.

Apply parsed annotation commands with identical behavior across all paths.

Apply theme-aware styles based on the active Qt palette.

Store pseudocode in the index cache with optional size limits. Returns (cached_count, skipped_count).

Verify Hex-Rays is loaded. Returns ``False`` with a warning if not.

Heuristic count of entry points for prompt context.

Create an OpenAI compatible client identically to annotator.py/gatherer.py

Drop oldest records with specified priority.

Return a fresh, empty analysis result dict.

Extract JSON from text that may contain additional content. Handles cases where LLM adds text like 'Here is the summary:' before JSON

Apply user toggles consistently before preview/apply.

Mark the current turn complete while preserving chat context for follow-up queries.

Format retrieved memories as prompt section.

Returns the shared client or creates a transient one if none exists.

Compute the standard index directory path for *identifier*.

Legacy index file path (before per-binary folders).

Resolve a cross-platform writable directory for Fanalysis logs.

Extract valid model IDs from MODEL_LIST, with default fallback.

Generate keyword-based summary without LLM.

Attempt to register a tag with the dynamic manager. Returns True if new.

Generate metadata-only summary.