Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Python agentaianthropicclaude-codecompressioncontext-engineeringcontext-windowcursorfastapilangchainllmmcpopenaiprompt-engineeringproxypythonragtoken-optimizationtokenstypescript
Integrate codebase-memory-mcp, disable CodeCompressor
chopratejasd8a4ed3dc99d
diff --git a/headroom/cli/wrap.py b/headroom/cli/wrap.pyindex c530398f0..3ef486ea9 100644--- a/headroom/cli/wrap.py+++ b/headroom/cli/wrap.py@@ -182,6 +182,79 @@ def _setup_rtk(verbose: bool = False) -> Path | None:return rtk_path+def _setup_code_graph(verbose: bool = False) -> bool:+ """Ensure codebase-memory-mcp is installed and project is indexed.++ codebase-memory-mcp builds a knowledge graph of the codebase using+ tree-sitter, enabling the LLM to query code structure (call chains,+ function definitions, impact analysis) instead of reading entire files.++ With Claude Code's MCP Tool Search, the 14 graph tools add ~200 tokens+ overhead per request (not the full ~1,915) — they're lazy-loaded.++ Returns True if graph is ready, False if setup failed.+ """+ cbm_bin = shutil.which("codebase-memory-mcp")+ if not cbm_bin:+ if verbose:+ click.echo(" Code graph: codebase-memory-mcp not found, skipping")+ click.echo(+ " Install: curl -fsSL https://raw.githubusercontent.com/"+ "DeusData/codebase-memory-mcp/main/install.sh | sh"+ )+ return False++ # Index current project (fast — ~1s for most repos, idempotent)+ project_dir = str(Path.cwd())+ try:+ result = subprocess.run(+ [+ cbm_bin,+ "cli",+ "index_repository",+ json.dumps({"repo_path": project_dir, "mode": "fast"}),+ ],+ capture_output=True,+ text=True,+ timeout=30,+ )+ if result.returncode == 0:+ # Parse node/edge counts from output+ for line in result.stdout.splitlines():+ if '"nodes"' in line:+ try:+ # Parse the JSON response to extract node/edge counts+ import re++ m_nodes = re.search(r'"nodes":(\d+)', line)+ m_edges = re.search(r'"edges":(\d+)', line)+ if m_nodes and m_edges:+ nodes = int(m_nodes.group(1))+ edges = int(m_edges.group(1))+ click.echo(+ f" Code graph: indexed ({nodes:,} symbols, "+ f"{edges:,} relationships)"+ )+ else:+ click.echo(" Code graph: indexed")+ except (ValueError, AttributeError):+ click.echo(" Code graph: indexed")+ return True+ click.echo(" Code graph: indexed")+ return True+ else:+ if verbose:+ click.echo(f" Code graph: indexing failed ({result.stderr[:100]})")+ return False+ except subprocess.TimeoutExpired:+ click.echo(" Code graph: indexing timed out (will complete in background)")+ return False+ except Exception as e:+ if verbose:+ click.echo(f" Code graph: setup failed ({e})")+ return False++# rtk instructions for tools without hook support (Codex, Cursor, Aider).# These get injected into AGENTS.md / .cursorrules so the LLM voluntarily# uses rtk-prefixed commands. Kept concise to minimize instruction overhead.@@ -464,6 +537,7 @@ def _launch_tool(*,learn: bool = False,agent_type: str = "unknown",+ no_code_graph: bool = False,backend: str | None = None,anyllm_provider: str | None = None,region: str | None = None,@@ -492,6 +566,9 @@ def _launch_tool(region=region,)+ if not no_code_graph:+ _setup_code_graph(verbose=False)+click.echo()click.echo(f" Launching {tool_label} (API routed through Headroom)...")for var in env_vars_display:@@ -779,6 +856,9 @@ def unwrap() -> None:@wrap.command(context_settings={"ignore_unknown_options": True})@click.option("--port", "-p", default=8787, type=int, help="Proxy port (default: 8787)")@click.option("--no-rtk", is_flag=True, help="Skip rtk installation and hook registration")+ "--no-code-graph", is_flag=True, help="Skip code graph indexing (codebase-memory-mcp)"+)@click.option("--no-proxy", is_flag=True, help="Skip proxy startup (use existing proxy)")@click.option("--learn", is_flag=True, help="Enable live traffic learning (patterns saved to MEMORY.md)"@@ -789,6 +869,7 @@ def unwrap() -> None:def claude(port: int,no_rtk: bool,+ no_code_graph: bool,no_proxy: bool,learn: bool,verbose: bool,@@ -803,11 +884,11 @@ def claude(\bExamples:- headroom wrap claude # Start everything- headroom wrap claude --resume <id> # Resume a session- headroom wrap claude -- -p # Claude in print mode- headroom wrap claude --port 9999 # Custom proxy port- headroom wrap claude --no-rtk # Skip rtk (proxy only)+ headroom wrap claude # Start everything+ headroom wrap claude --resume <id> # Resume a session+ headroom wrap claude -- -p # Claude in print mode+ headroom wrap claude --no-code-graph # Skip code graph+ headroom wrap claude --no-rtk # Skip rtk (proxy only)"""if prepare_only:if not no_rtk:@@ -841,6 +922,11 @@ def claude(elif verbose:click.echo(" Skipping rtk (--no-rtk)")+ if not no_code_graph:+ _setup_code_graph(verbose=verbose)+ elif verbose:+ click.echo(" Skipping code graph (--no-code-graph)")+click.echo()click.echo(" Launching Claude Code (API routed through Headroom)...")click.echo(f" ANTHROPIC_BASE_URL=http://127.0.0.1:{port}")@@ -1019,6 +1105,9 @@ def copilot(@wrap.command(context_settings={"ignore_unknown_options": True})@click.option("--port", "-p", default=8787, type=int, help="Proxy port (default: 8787)")@click.option("--no-rtk", is_flag=True, help="Skip rtk installation and AGENTS.md injection")+ "--no-code-graph", is_flag=True, help="Skip code graph indexing (codebase-memory-mcp)"+)@click.option("--no-proxy", is_flag=True, help="Skip proxy startup (use existing proxy)")@click.option("--learn", is_flag=True, help="Enable live traffic learning (patterns saved to AGENTS.md)"@@ -1042,6 +1131,7 @@ def copilot(def codex(port: int,no_rtk: bool,+ no_code_graph: bool,no_proxy: bool,learn: bool,backend: str | None,@@ -1107,6 +1197,7 @@ def codex(env_vars_display=[f"OPENAI_BASE_URL=http://127.0.0.1:{port}/v1"],learn=learn,agent_type="codex",+ no_code_graph=no_code_graph,backend=backend,anyllm_provider=anyllm_provider,region=region,@@ -1121,6 +1212,9 @@ def codex(@wrap.command(context_settings={"ignore_unknown_options": True})@click.option("--port", "-p", default=8787, type=int, help="Proxy port (default: 8787)")@click.option("--no-rtk", is_flag=True, help="Skip rtk installation and conventions injection")+ "--no-code-graph", is_flag=True, help="Skip code graph indexing (codebase-memory-mcp)"+)@click.option("--no-proxy", is_flag=True, help="Skip proxy startup (use existing proxy)")@click.option("--learn", is_flag=True, help="Enable live traffic learning")@click.option(@@ -1134,6 +1228,7 @@ def codex(def aider(port: int,no_rtk: bool,+ no_code_graph: bool,no_proxy: bool,learn: bool,backend: str | None,@@ -1193,6 +1288,7 @@ def aider(],learn=learn,agent_type="aider",+ no_code_graph=no_code_graph,backend=backend,anyllm_provider=anyllm_provider,region=region,diff --git a/headroom/proxy/models.py b/headroom/proxy/models.pyindex c9a837ac7..cb9f37834 100644--- a/headroom/proxy/models.py+++ b/headroom/proxy/models.py@@ -110,8 +110,8 @@ class ProxyConfig:ccr_proactive_expansion: bool = Trueccr_max_proactive_expansions: int = 2- # Code-aware compression- code_aware_enabled: bool = True+ # Code-aware compression (disabled by default — use code graph tools instead)+ code_aware_enabled: bool = False# Per-tool compression profilestool_profiles: dict[str, Any] | None = Nonediff --git a/headroom/transforms/content_router.py b/headroom/transforms/content_router.pyindex 7e4399a7b..c7c051887 100644--- a/headroom/transforms/content_router.py+++ b/headroom/transforms/content_router.py@@ -388,7 +388,7 @@ class ContentRouterConfig:"""# Enable/disable specific compressors- enable_code_aware: bool = True+ enable_code_aware: bool = False # Disabled: use code graph MCP tools insteadenable_kompress: bool = True # Kompress: ModernBERT token compressorenable_smart_crusher: bool = Trueenable_search_compressor: bool = True@@ -397,7 +397,7 @@ class ContentRouterConfig:enable_image_optimizer: bool = True # Image token optimization# Routing preferences- prefer_code_aware_for_code: bool = True+ prefer_code_aware_for_code: bool = False # Disabled: let code pass through unmangledmixed_content_threshold: int = 2 # Min types to consider mixedmin_section_tokens: int = 20 # Min tokens to compress a section