Broaden filter: drop non-conversational classifier calls (max_tokens<=64, no tools) — removes residual OSL 5-10 spike

README.md

@@ -46,18 +46,21 @@ Filters:
    is on top of that.
 3. **CC ≥ 2.1.139.** Every replayable request in every included session
    originated from a Claude Code CLI build at version 2.1.139 or newer.
-4. **Security Monitor excluded.** Claude Code's "Security Monitor"
-   subagent fires short low-OSL (5-10 token) classification calls. These
-   are dropped at source so they don't pollute the OSL distribution
-   (`subagent_label IS DISTINCT FROM 'Security Monitor'` plus a body-
-   pattern fallback for Security Monitor rows that got the generic
-   'Subagent' label).
+4. **Non-conversational classifier calls excluded.** Claude Code fires
+   several short auxiliary classifier calls per session ("SUGGESTION
+   MODE" next-input prediction, conversation-title generation, haiku
+   intent detection, the "Security Monitor" subagent). They share a
+   clean request shape — `max_tokens <= 64` AND no `tools` — and
+   produced a heavy OSL 5-10 spike (~9k+ rows at exactly OSL=7) that
+   pollutes the published output-length distribution. They are dropped
+   at source by shape, with an additional label/body fallback for the
+   Security Monitor subagent specifically.
 
 - **Traces:** 219
-- **Top-level entries:** 48,836 (46,896 main turns + 1,940 subagent groups)
-- **Sub-agent inner requests:** 39,251
-- **Total individual model requests:** 86,147
-- **Models:** `claude-opus-4-7` (82,809 req), `claude-opus-4-6` (2,138 req), `claude-haiku-4-5-20251001` (1,104 req), `claude-sonnet-4-6` (96 req)
+- **Top-level entries:** 39,726 (37,732 main turns + 1,994 subagent groups)
+- **Sub-agent inner requests:** 40,367
+- **Total individual model requests:** 78,099
+- **Models:** `claude-opus-4-7` (74,293 req), `claude-opus-4-6` (2,138 req), `claude-haiku-4-5-20251001` (1,163 req), `claude-sonnet-4-6` (505 req)
 - **KV block size:** 64 tokens
 - **Hash scope:** `local` — block hash IDs are only comparable *within*
   a single trace; they are not a global content-addressable identity.
@@ -129,19 +132,19 @@ triggered them.
 
 ## Summary statistics
 
-Across **all** requests (main + sub-agent inner = 86,147 requests):
+Across **all** requests (main + sub-agent inner = 78,099 requests):
 
 |                              |   p50    |   p75    |   p90    |   p95    |   p99    |   mean   |
 |------------------------------|---------:|---------:|---------:|---------:|---------:|---------:|
-| ISL (tokens)                 |   98,231 |  197,519 |  373,531 |  511,098 |  779,719 |  154,048 |
-| OSL (tokens)                 |      263 |      598 |    1,336 |    2,032 |    6,408 |      616 |
-| Top entries/trace            |      139 |      272 |      497 |      722 |    1,114 |      222 |
+| ISL (tokens)                 |  109,729 |  214,992 |  394,994 |  535,364 |  799,788 |  164,194 |
+| OSL (tokens)                 |      307 |      688 |    1,442 |    2,204 |    6,852 |      694 |
+| Top entries/trace            |      114 |      219 |      369 |      543 |    1,062 |      181 |
 
 ## Plots
 
 ### Main-agent stream
 
-Histograms across all 46,896 main-agent turns (sub-agent groups skipped).
+Histograms across all 37,732 main-agent turns (sub-agent groups skipped).
 Same view as the [no-subagents sibling](https://huggingface.co/datasets/semianalysisai/cc-traces-weka-no-subagents-051826) — the two datasets share an identical main-agent stream.
 
 ![Main-stream distributions — log x](plots/distributions_log.png)
@@ -149,7 +152,7 @@ Same view as the [no-subagents sibling](https://huggingface.co/datasets/semianal
 
 ### Sub-agent fan-out analysis
 
-Histograms across all 1,940 sub-agent groups (and their 39,251 inner requests). Companion to the main-stream plots — answers questions about how often sub-agents are spawned, how deep their tool-loops go, how long they take, and how much intra-group prefix cache reuse exists.
+Histograms across all 1,994 sub-agent groups (and their 40,367 inner requests). Companion to the main-stream plots — answers questions about how often sub-agents are spawned, how deep their tool-loops go, how long they take, and how much intra-group prefix cache reuse exists.
 
 ![Sub-agent distributions — log x](plots/subagent_distributions_log.png)
 ![Sub-agent distributions — linear x](plots/subagent_distributions_linear.png)
@@ -167,10 +170,10 @@ Six panels:
 
 | model | requests (main + subagent inner) |
 |-------|---------:|
-| `claude-opus-4-7` | 82,809 |
+| `claude-opus-4-7` | 74,293 |
 | `claude-opus-4-6` | 2,138 |
-| `claude-haiku-4-5-20251001` | 1,104 |
-| `claude-sonnet-4-6` | 96 |
+| `claude-haiku-4-5-20251001` | 1,163 |
+| `claude-sonnet-4-6` | 505 |
 
 ## Source
 
@@ -179,8 +182,10 @@ Same proxy → weka pipeline as the
 just without the post-step that strips `WekaSubagentEntry` blocks.
 
   1. `utils/sample_proxy_traces.py --min-trace-version 5 --min-main-turns 20 --require-cli-min 2.1.139 --privacy-mode anon`
-     (Security Monitor exclusion is hardcoded into the query so the
-     5-10-token OSL spike doesn't leak into the published corpus.)
+     (Non-conversational classifier-call exclusion — `max_tokens <= 64`
+     + no `tools` array — is hardcoded into the query so the 5-10-token
+     OSL spike doesn't leak into the published corpus. A label/body
+     fallback for the Security Monitor subagent is layered on top.)
   2. `utils/proxy_to_weka.py` (subagent grouping per the dashboard's
      algorithm)
   3. Concatenate the resulting weka JSONs into `traces.jsonl` — one trace

traces.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:69d08f80d3abbc1a83b6f9ec71a3b1e85a79c9996e8c8fe0ce70e8cbde8dd159
-size 1326859564
+oid sha256:8e0f7cba9bbd40fc28af41bccc8afccf839bc77d081fb11c210a2939eb1215f5
+size 1076644069