Session Benchmark v3

v3: Performance profiling of session query, search, highlight, and cross-reference operations

iTerm Panes

JSONL Files

353

JSONL Total

627 MB

Trials/Op

Single Operations

Operation	Min (ms)	Median (ms)	Max (ms)	Stdev	Distribution	Detail
session_tool list (alive, json)	770	785	860	35	785ms	17 lines output
session_tool cleanup (dry-run)	383	408	412	14	408ms	44 lines output
rg -c 'model' (-Users-tchklovski-all-code-rivus/)	50	58	354	133	58ms	2013 lines output
it2api get-buffer (single)	300	307	313	5	307ms	40 lines output
tab color set+unset cycle	906	914	978	29	914ms
highlight flow (1 panes)	1203	1222	1263	23	1222ms
cross-ref sessions.yaml ↔ it2api	289	300	310	8	300ms	1 alive, 42 stale of 43
find --state idle	388	394	423	16	394ms	5 lines output
find 'model' (content search)	749	768	1118	158	768ms	5 lines output
find 'benchmark' --all	724	745	783	22	745ms	26 lines output

Find Command Performance

session find uses a 3-phase search: (1) metadata match — badge + name, (2) rg content search across JSONL, (3) parallel buffer reads for state enrichment. Phases are skipped when not needed (e.g., state-only queries skip rg entirely).

Query	Goal	Phases Used	Min (ms)	Median (ms)	Max (ms)	Bar
`find --state idle`	Find all idle sessions	metadata only	388	394	423	394ms
`find 'model' (content search)`	Keyword search across conversations	metadata + rg	749	768	1118	768ms
`find 'benchmark' --all`	Search including dead sessions	metadata + rg (all)	724	745	783	745ms

Find Query Anatomy

Phase	What	Cost	When
1. Metadata	Load sessions.yaml + it2api list-sessions + hub badges	~320ms	Always
2. Content	rg across 620MB JSONL	~60ms	Only if query provided
3. Enrich	Parallel it2api get-buffer for alive matches	N × 320ms / threads	Only if state filter or permission detection needed

With N panes: Phase 3 is the parallelism opportunity. Sequential: N × 320ms. ThreadPool(8): ceil(N/8) × 320ms. MCPretentious: N × 1.2ms.

Scenario	1 pane	5 panes	15 panes
Sequential it2api	320ms	1,600ms	4,800ms
ThreadPool(8)	320ms	320ms	640ms
MCPretentious	1ms	6ms	18ms

Parallelism Comparison (Sequential vs Concurrent)

Operation	Sequential (ms)	Concurrent (ms)	Speedup	Visual
buffer reads x1	290	306	0.95x	seq 290ms conc 306ms
tab color x1	936	980	0.96x	seq 936ms conc 980ms
rg x4 patterns	207	160	1.29x	seq 207ms conc 160ms

session_tool list Breakdown (Why is it 14+ seconds?)

Each component of get_enriched_sessions() measured independently. The total explains the ~14s wall time of session_tool list.

Component	Median (ms)	% of Total	Waterfall
claude --version	52	4.1%	52ms
it2api list-sessions	290	22.8%	290ms
JSONL rglob+version (1 session)	13	1.0%	13ms
JSONL rglob+version (43 sessions)	615	48.5%	615ms
it2api get-buffer x1 seq (all alive)	298	23.5%	298ms
Hub DB badge load	1	0.1%	1ms
JSONL wait state detect (1 session)	1	0.0%	1ms
SUM (estimated)	1269	100%

All Results (raw)

Operation	Category	Min	Median	Mean	Max	Stdev	Trials
session_tool list (alive, json)	sequential	770	785	802	860	35	5
session_tool cleanup (dry-run)	sequential	383	408	399	412	14	5
rg -c 'model' (-Users-tchklovski-all-code-rivus/)	sequential	50	58	118	354	133	5
it2api get-buffer (single)	sequential	300	307	308	313	5	5
tab color set+unset cycle	sequential	906	914	929	978	29	5
highlight flow (1 panes)	sequential	1203	1222	1228	1263	23	5
cross-ref sessions.yaml ↔ it2api	sequential	289	300	301	310	8	5
find --state idle	sequential	388	394	402	423	16	5
find 'model' (content search)	sequential	749	768	837	1118	158	5
find 'benchmark' --all	sequential	724	745	747	783	22	5
buffer reads x1 (sequential)	sequential	287	290	293	307	8	5
buffer reads x1 (concurrent)	concurrent	294	306	307	330	14	5
tab color x1 (sequential)	sequential	933	936	945	973	17	5
tab color x1 (concurrent)	concurrent	948	980	973	985	16	5
rg x4 patterns (sequential)	sequential	200	207	220	254	23	5
rg x4 patterns (concurrent)	concurrent	159	160	161	163	2	5

Optimization Roadmap

Current Bottleneck: it2api subprocess overhead (~300ms/call)

Each it2api call spawns a Python process. With 15+ alive panes needing buffer reads, that's 15 × 305ms = 4.5s just in process startup overhead.

Projected: session_tool list latency

Scenario	Buffer Reads	JSONL Version	list-sessions	Other	Total	vs Current
Current (it2api, sequential)	298ms	615ms	290ms	62ms	1265ms	baseline
+ ThreadPool parallelism	99ms	615ms	290ms	62ms	1066ms	1.2x
+ MCPretentious (255x buffer)	1ms	615ms	145ms	62ms	823ms	1.5x
+ direct JSONL path (no rglob)	1ms	50ms	145ms	62ms	258ms	5x

Action Items (by impact)

#	Change	Saves	Effort
1	MCPretentious for get-buffer + send-text Replace it2api subprocess calls in supervisor/adapters/iterm2.py with persistent WebSocket	~297ms (255x)	medium
2	Direct JSONL path instead of rglob `JSONL_DIR / f"{sid}.jsonl"` instead of `rglob(f"{sid}.jsonl")`	~565ms	trivial
3	ThreadPool for remaining it2api calls Parallelize badge/title/profile property calls that can't use MCPretentious	~199ms (if no MCP)	easy
4	Cache claude --version Version doesn't change within a session — cache for 5 min	~52ms	trivial

Bottleneck Analysis

Primary bottleneck: it2api calls

it2api operations median: 611ms — each call has ~300ms Python startup overhead
Disk operations (rg) median: 160ms — rg is fast, rglob is slow
session_tool / cross-ref median: 408ms

Parallelism Gains (measured)

buffer reads x1: 0.95x speedup (290ms → 306ms)
tab color x1: 0.96x speedup (936ms → 980ms)
rg x4 patterns: 1.29x speedup (207ms → 160ms)

MCPretentious Gains (from evaluation)

Read screen: 305ms → 1.2ms = 255x speedup
Send text: 296ms → 1.0ms = 296x speedup
List sessions: 300ms → 156ms = 2x speedup
Mechanism: persistent WebSocket + protobuf vs subprocess spawn per call

Ranked by Latency (slowest first)

highlight flow (1 panes): 1222ms median
tab color x1 (concurrent): 980ms median
tab color x1 (sequential): 936ms median
tab color set+unset cycle: 914ms median
session_tool list (alive, json): 785ms median
find 'model' (content search): 768ms median
find 'benchmark' --all: 745ms median
session_tool cleanup (dry-run): 408ms median
find --state idle: 394ms median
it2api get-buffer (single): 307ms median

Generated: 2026-02-06 11:38:27