coding-agent-cost 0.2.1__tar.gz → 0.3.0__tar.gz

{coding_agent_cost-0.2.1/coding_agent_cost.egg-info → coding_agent_cost-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: coding-agent-cost
-Version: 0.2.1
+Version: 0.3.0
 Summary: Estimate AI coding agent (Claude Code / Codex CLI) token usage and cost from local logs
 Author: shiki-yusuke
 License: MIT
@@ -73,7 +73,8 @@ agent-cost report
 The PyPI distribution is named [`coding-agent-cost`](https://pypi.org/project/coding-agent-cost/),
 while the command remains `agent-cost` and the import remains `agent_cost`. The shorter PyPI
 name is unavailable because of PyPI's similarity rule; this project is not affiliated with the
-unrelated `agentcost` distribution.
+unrelated `agentcost` distribution. Releases are published from a `v*` tag by GitHub Actions
+through PyPI's Trusted Publisher; see [docs/release.md](docs/release.md).
 
 ## Synthetic output example
 
@@ -231,6 +232,14 @@ network, never calls `gh`, and never resolves branches or PRs.
   flagged, since the reader's own cross-check of real transcripts found
   exactly that pattern in every observed duplicated group. See
   `CHANGELOG.md`'s 0.2.0 entry.
+  If the row a group adopts has no `stop_reason` (common in subagent
+  transcripts, where a message ending in `tool_use` may never get its final
+  line), its `output_tokens` is only the streaming head's value: the
+  group's `output` fact is flagged `source_quality: "output_lower_bound"`
+  -- the observed lower bound of that message's output tokens, not
+  guaranteed to be >= the true count. The group's input-side facts stay
+  `"ok"`, and the token amount itself is unchanged (still priced and
+  included in rows/totals); the flag is a warning only.
   When Anthropic's prompt-cache TTL breakdown (5-minute vs 1-hour writes) is
   present in the log, it's used; otherwise the cache-write tokens are priced
   at the 5-minute rate as an explicit **lower bound** and flagged
@@ -341,7 +350,7 @@ agent-cost measure --session-id <id> [--session-id <id> ...] \
 ```json
 {
   "protocol_version": "measure/v1",
-  "producer_version": "0.2.1",
+  "producer_version": "0.3.0",
   "accounting_basis": "agent-cost-raw-total/v2",
   "generated_at": "...",
   "window": { "since": "...", "until": null },
@@ -362,7 +371,7 @@ agent-cost measure --session-id <id> [--session-id <id> ...] \
     "duplicate_rows_skipped": 0,
     "conflicting_duplicate_groups": 0,
     "missing_dedup_identity_rows": 0,
-    "source_quality": { "ok": 41, "first_event_delta": 2, "identity_missing": 0 }
+    "source_quality": { "ok": 41, "first_event_delta": 2, "identity_missing": 0, "output_lower_bound": 1 }
   }
 }
 ```
@@ -394,8 +403,9 @@ includes absolute file paths, rollout paths, prompt/message content, or git
 branch names -- only the fields needed to reproduce a cost estimate:
 `occurred_at_utc`, `agent`, `session_id`, `model_raw`, `model_key`,
 `token_kind`, `tokens`, `mode`, and `source_quality` (a fixed-vocabulary
-caveat about how that one fact was derived, e.g. `"ok"` or Codex's
-`"first_event_delta"` -- never null).
+caveat about how that one fact was derived: `"ok"`, Codex's
+`"first_event_delta"`, or Claude's `"identity_missing"` /
+`"output_lower_bound"` -- never null).
 
 ## License
 

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/README.md

@@ -47,7 +47,8 @@ agent-cost report
 The PyPI distribution is named [`coding-agent-cost`](https://pypi.org/project/coding-agent-cost/),
 while the command remains `agent-cost` and the import remains `agent_cost`. The shorter PyPI
 name is unavailable because of PyPI's similarity rule; this project is not affiliated with the
-unrelated `agentcost` distribution.
+unrelated `agentcost` distribution. Releases are published from a `v*` tag by GitHub Actions
+through PyPI's Trusted Publisher; see [docs/release.md](docs/release.md).
 
 ## Synthetic output example
 
@@ -205,6 +206,14 @@ network, never calls `gh`, and never resolves branches or PRs.
   flagged, since the reader's own cross-check of real transcripts found
   exactly that pattern in every observed duplicated group. See
   `CHANGELOG.md`'s 0.2.0 entry.
+  If the row a group adopts has no `stop_reason` (common in subagent
+  transcripts, where a message ending in `tool_use` may never get its final
+  line), its `output_tokens` is only the streaming head's value: the
+  group's `output` fact is flagged `source_quality: "output_lower_bound"`
+  -- the observed lower bound of that message's output tokens, not
+  guaranteed to be >= the true count. The group's input-side facts stay
+  `"ok"`, and the token amount itself is unchanged (still priced and
+  included in rows/totals); the flag is a warning only.
   When Anthropic's prompt-cache TTL breakdown (5-minute vs 1-hour writes) is
   present in the log, it's used; otherwise the cache-write tokens are priced
   at the 5-minute rate as an explicit **lower bound** and flagged
@@ -315,7 +324,7 @@ agent-cost measure --session-id <id> [--session-id <id> ...] \
 ```json
 {
   "protocol_version": "measure/v1",
-  "producer_version": "0.2.1",
+  "producer_version": "0.3.0",
   "accounting_basis": "agent-cost-raw-total/v2",
   "generated_at": "...",
   "window": { "since": "...", "until": null },
@@ -336,7 +345,7 @@ agent-cost measure --session-id <id> [--session-id <id> ...] \
     "duplicate_rows_skipped": 0,
     "conflicting_duplicate_groups": 0,
     "missing_dedup_identity_rows": 0,
-    "source_quality": { "ok": 41, "first_event_delta": 2, "identity_missing": 0 }
+    "source_quality": { "ok": 41, "first_event_delta": 2, "identity_missing": 0, "output_lower_bound": 1 }
   }
 }
 ```
@@ -368,8 +377,9 @@ includes absolute file paths, rollout paths, prompt/message content, or git
 branch names -- only the fields needed to reproduce a cost estimate:
 `occurred_at_utc`, `agent`, `session_id`, `model_raw`, `model_key`,
 `token_kind`, `tokens`, `mode`, and `source_quality` (a fixed-vocabulary
-caveat about how that one fact was derived, e.g. `"ok"` or Codex's
-`"first_event_delta"` -- never null).
+caveat about how that one fact was derived: `"ok"`, Codex's
+`"first_event_delta"`, or Claude's `"identity_missing"` /
+`"output_lower_bound"` -- never null).
 
 ## License
 

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/agent_cost/__init__.py

@@ -6,4 +6,4 @@ against a versioned rate catalog. Everything happens locally; there are no
 network calls.
 """
 
-__version__ = "0.2.1"
+__version__ = "0.3.0"

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/agent_cost/facts.py

@@ -32,10 +32,14 @@ MODES = ("fast", "normal", "unknown")
 # (e.g. Codex's first delta in a rollout is measured against an assumed
 # zero baseline; Claude's reader can't dedup a row that lacks a full
 # (message.id, requestId) pair, so it emits that row individually and
-# flags it "identity_missing" rather than silently treating it as "ok").
+# flags it "identity_missing" rather than silently treating it as "ok";
+# "output_lower_bound" marks a Claude output fact whose tokens are only the
+# observed lower bound of that message's output_tokens -- the reader's
+# adopted row had no stop_reason, so the value is not guaranteed to be
+# >= the true count).
 # This is never left unset -- every Fact defaults to "ok" so export never
 # emits a null source_quality.
-SOURCE_QUALITY_VALUES = ("ok", "first_event_delta", "identity_missing")
+SOURCE_QUALITY_VALUES = ("ok", "first_event_delta", "identity_missing", "output_lower_bound")
 
 _BRACKET_SUFFIX = re.compile(r"\[[^\]]*\]$")
 _DATE_SUFFIX = re.compile(r"@\d{6,8}$")

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/agent_cost/rates.json

@@ -1,6 +1,6 @@
 {
   "schema_version": "1",
-  "catalog_version": "2026-09-23",
+  "catalog_version": "2026-09-29",
   "currency": "USD",
   "unit": "per_mtok",
   "usd_per_credit": "0.04",
@@ -11,7 +11,8 @@
     "gpt-6-astra: Codex standard token-based rates only; not API pricing, legacy message metering, or actual contractual charges. Confirmed 2026-09-06 JST (2026-09-05T17:18:23Z). effective_from is this observation cutoff, NOT an official launch or price-start time; earlier events remain unpriced. No Codex long-context surcharge or cache-write charge is applied. Fast multiplier 2.5 applies to explicit priority request settings with matching owned turn/model context; explicit default uses Standard. Missing or ambiguous Astra modes remain unpriced. These settings do not confirm the backend processing tier. See docs/astra-pricing.md for compatibility limits and the separate GPT-5.6 discrepancy report.",
     "claude-fable-5-1 added 2026-09-09: rate_id claude-fable-5-1-local-first-observed-2026-09-02. effective_from (2026-09-02T01:17:55Z) is the timestamp the raw model ID was first observed across all 54 local transcript files (none found in August), not a confirmed official launch date -- if an authoritative launch date is confirmed later, add it as a separate rate_id with effective_until on this one rather than editing this period in place. Values (cache_read $0.25/MTok = 0.025x base input) differ from claude-fable-5's cache_read ($1.00/MTok = 0.1x); this is a genuine per-model price difference confirmed against the primary source (see sources below), not an inconsistency to reconcile.",
     "claude-sonnet-5 corrected 2026-09-09: a prior period (rate_id claude-sonnet-5-standard, effective_from 2026-09-01, values 3.0/0.30/3.75/6.0/15.0) priced a previously-announced rate increase that the pricing page's claude-sonnet-5-introductory-pricing note states did not occur (\"The previously scheduled increase to $3/$15 ... on September 1, 2026 will not occur\"). That period has been replaced by claude-sonnet-5-standard-2026-09-01, which carries the same $2/$0.20/$2.50/$4/$10 values as the preceding claude-sonnet-5-launch-promo period, since the launch price is confirmed to be the ongoing price with no change at the 2026-09-01 boundary. Any measure/v1 or report output computed against the pre-correction catalog for claude-sonnet-5 sessions occurring on or after 2026-09-01 overstated cost by 1.5x for that model. To detect this class of error going forward, cross-check each priced model's rate periods against its primary pricing page's Note text on a recurring (e.g. monthly) basis for mentions of a rate-change announcement being withdrawn or not taking effect -- a period can be internally valid (schema-wise) and still price a rate that was announced but never actually charged.",
-    "claude-opus-5-5 added 2026-09-23: rate_id claude-opus-5-5-launch-2026-09-22. effective_from 2026-09-22T00:00:00Z is 00:00:00Z of the official public launch date (anthropic.com/claude-opus-5-5, 'September 22, 2026'); the launch time of day is not published, but a transcript row whose raw model ID is claude-opus-5-5 cannot predate the launch, so pricing from the start of that UTC day cannot price a pre-launch event. Values 4.0/0.20/5.0/8.0/20.0 per the pricing page: cache_read is 0.05x base input (pricing-page footnote 2), a genuine per-model difference from the standard 0.1x and from claude-fable-5-1's 0.025x. fast_multiplier 2.0 (fast mode $8/$40 per MTok on the pricing page). Claude Code's CHANGELOG entry for 2.1.280 states Opus 5.5 became the default model (see sources); rows whose raw model ID is claude-opus-5-5 would be unpriced without this entry. Claude Sonnet 5.5 and Claude Haiku 5.5 are announced to follow 'in the coming weeks' -- add them as separate entries when their model IDs and prices are published, never as aliases of the 5.x entries."
+    "claude-opus-5-5 added 2026-09-23: rate_id claude-opus-5-5-launch-2026-09-22. effective_from 2026-09-22T00:00:00Z is 00:00:00Z of the official public launch date (anthropic.com/claude-opus-5-5, 'September 22, 2026'); the launch time of day is not published, but a transcript row whose raw model ID is claude-opus-5-5 cannot predate the launch, so pricing from the start of that UTC day cannot price a pre-launch event. Values 4.0/0.20/5.0/8.0/20.0 per the pricing page: cache_read is 0.05x base input (pricing-page footnote 2), a genuine per-model difference from the standard 0.1x and from claude-fable-5-1's 0.025x. fast_multiplier 2.0 (fast mode $8/$40 per MTok on the pricing page). Claude Code's CHANGELOG entry for 2.1.280 states Opus 5.5 became the default model (see sources); rows whose raw model ID is claude-opus-5-5 would be unpriced without this entry. Claude Sonnet 5.5 and Claude Haiku 5.5 are announced to follow 'in the coming weeks' -- add them as separate entries when their model IDs and prices are published, never as aliases of the 5.x entries.",
+    "claude-sonnet-5-5 added 2026-09-29: rate_id claude-sonnet-5-5-launch-2026-09-28. effective_from 2026-09-28T00:00:00Z is 00:00:00Z of the official public launch date (anthropic.com/claude-sonnet-5-5, 'September 28, 2026'); as with claude-opus-5-5, the launch time of day is not published, but a transcript row whose raw model ID is claude-sonnet-5-5 cannot predate the launch, so pricing from the start of that UTC day cannot price a pre-launch event. Values 2.0/0.20/2.50/4.0/10.0 per the pricing page (Sonnet 5.5 row, no footnote: cache_read is the standard 0.1x, unlike claude-opus-5-5's 0.05x and claude-fable-5-1's 0.025x). Identical to claude-sonnet-5's ongoing values, but kept as a separate entry, never an alias: the pricing page lists the two rows independently and a future change to one need not apply to the other. fast_multiplier 1.0: the pricing page's fast-mode table lists only the Opus models. Claude Code's CHANGELOG entry for 2.1.284 states Sonnet 5.5 is now the default Sonnet model, so the `sonnet` alias in agent frontmatter resolves to it and rows whose raw model ID is claude-sonnet-5-5 are expected from that version onward; no such row existed in the local transcripts on 2026-09-29. Claude Haiku 5.5 is announced to follow 'in the coming weeks' -- add it as a separate entry when its model ID and prices are published."
   ],
   "sources": [
     {
@@ -81,6 +82,21 @@
       "url": "https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md",
       "retrieved_at": "2026-09-23",
       "note": "Claude Code 2.1.280 entry: Claude Opus 5.5 became the default model (1M context, $4/$20 per MTok, $0.20/MTok cache reads); the reason claude-opus-5-5 rows are expected in transcripts from that version"
+    },
+    {
+      "url": "https://platform.claude.com/docs/en/about-claude/pricing",
+      "retrieved_at": "2026-09-29",
+      "note": "used for claude-sonnet-5-5 (Sonnet 5.5 row: $2 / $2.50 / $4 / $0.20 / $10, no footnote so cache_read is the standard 0.1x; fast-mode table lists Opus models only); fetched via WebFetch, no local snapshot"
+    },
+    {
+      "url": "https://www.anthropic.com/claude-sonnet-5-5",
+      "retrieved_at": "2026-09-29",
+      "note": "Sonnet 5.5 announcement: public launch date September 28, 2026, model ID claude-sonnet-5-5, $2/$10 per MTok, Haiku 5.5 to follow in the coming weeks"
+    },
+    {
+      "url": "https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md",
+      "retrieved_at": "2026-09-29",
+      "note": "Claude Code 2.1.284 entry: added Claude Sonnet 5.5 (claude-sonnet-5-5), now the default Sonnet model; the reason claude-sonnet-5-5 rows are expected in transcripts from that version"
     }
   ],
   "models": [
@@ -264,6 +280,23 @@
         }
       ]
     },
+    {
+      "model_key": "claude-sonnet-5-5",
+      "aliases": [],
+      "fast_multiplier": "1.0",
+      "rates": [
+        {
+          "rate_id": "claude-sonnet-5-5-launch-2026-09-28",
+          "effective_from": "2026-09-28T00:00:00+00:00",
+          "effective_until": null,
+          "input_nocache": "2.0",
+          "cache_read": "0.20",
+          "cache_write_5m": "2.50",
+          "cache_write_1h": "4.0",
+          "output": "10.0"
+        }
+      ]
+    },
     {
       "model_key": "claude-haiku-4-5",
       "aliases": [],

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/agent_cost/readers/claude.py

@@ -278,6 +278,14 @@ def _resolve_mode(rows: list, adopted: dict) -> str:
     return adopted["mode"]
 
 
+def _is_final(row: dict) -> bool:
+    """Whether a row carries a real ``message.stop_reason`` -- a non-empty
+    ``str``. Null, missing, ``False``, ``""`` and non-string values all
+    fail to qualify."""
+    stop_reason = row["stop_reason"]
+    return isinstance(stop_reason, str) and bool(stop_reason)
+
+
 def _select_adopted_row(rows: list) -> dict:
     """Pick the row within one dedup group whose usage gets emitted.
 
@@ -302,8 +310,7 @@ def _select_adopted_row(rows: list) -> dict:
     """
     adopted_idx = None
     for idx in range(len(rows) - 1, -1, -1):
-        stop_reason = rows[idx]["stop_reason"]
-        if isinstance(stop_reason, str) and stop_reason:
+        if _is_final(rows[idx]):
             adopted_idx = idx
             break
 
@@ -366,6 +373,19 @@ def parse_session_detailed(jsonl_path: Path) -> ClaudeParseResult:
     the group has exactly one concrete mode elsewhere, that concrete mode
     is emitted instead (see its docstring).
 
+    If the adopted row itself is not final (no non-empty-string
+    ``stop_reason`` -- see ``_is_final``), the group's ``output`` fact gets
+    ``source_quality="output_lower_bound"``: that row's ``output_tokens``
+    is a streaming head's value, observed to be a lower bound on the
+    message's real output count (subagent transcripts often end a
+    tool_use message without ever writing its final row). This keys off
+    the *adopted* row, not "does the group contain a final row", so a
+    final row overridden by a later, differing non-final row is flagged
+    too. The group's input-side facts stay ``"ok"`` (those fields don't
+    grow while streaming), and token amounts are emitted unchanged --
+    this only marks the caveat. Identity-missing rows never form a group
+    and keep ``"identity_missing"``.
+
     ``occurred_at_utc`` on the emitted fact is the *adopted* row's own
     timestamp (not the group's first-seen timestamp): the adopted row is
     what determines the actual token counts, and pricing/month-bucketing
@@ -493,6 +513,7 @@ def parse_session_detailed(jsonl_path: Path) -> ClaudeParseResult:
         if marker in standalone_rows:
             row = standalone_rows[marker]
             source_quality = "identity_missing"
+            output_source_quality = source_quality
             dedup_units.append(
                 ClaudeDedupUnit(
                     session_id=row["sid"],
@@ -544,6 +565,7 @@ def parse_session_detailed(jsonl_path: Path) -> ClaudeParseResult:
                 "tokens": adopted["tokens"],
             }
             source_quality = "ok"
+            output_source_quality = "ok" if _is_final(adopted) else "output_lower_bound"
 
         model_key = normalize_model_key(row["model_raw"])
         for kind, amount in row["tokens"].items():
@@ -557,7 +579,7 @@ def parse_session_detailed(jsonl_path: Path) -> ClaudeParseResult:
                     token_kind=kind,
                     tokens=amount,
                     mode=row["mode"],
-                    source_quality=source_quality,
+                    source_quality=output_source_quality if kind == "output" else source_quality,
                 )
             )
 

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0/coding_agent_cost.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: coding-agent-cost
-Version: 0.2.1
+Version: 0.3.0
 Summary: Estimate AI coding agent (Claude Code / Codex CLI) token usage and cost from local logs
 Author: shiki-yusuke
 License: MIT
@@ -73,7 +73,8 @@ agent-cost report
 The PyPI distribution is named [`coding-agent-cost`](https://pypi.org/project/coding-agent-cost/),
 while the command remains `agent-cost` and the import remains `agent_cost`. The shorter PyPI
 name is unavailable because of PyPI's similarity rule; this project is not affiliated with the
-unrelated `agentcost` distribution.
+unrelated `agentcost` distribution. Releases are published from a `v*` tag by GitHub Actions
+through PyPI's Trusted Publisher; see [docs/release.md](docs/release.md).
 
 ## Synthetic output example
 
@@ -231,6 +232,14 @@ network, never calls `gh`, and never resolves branches or PRs.
   flagged, since the reader's own cross-check of real transcripts found
   exactly that pattern in every observed duplicated group. See
   `CHANGELOG.md`'s 0.2.0 entry.
+  If the row a group adopts has no `stop_reason` (common in subagent
+  transcripts, where a message ending in `tool_use` may never get its final
+  line), its `output_tokens` is only the streaming head's value: the
+  group's `output` fact is flagged `source_quality: "output_lower_bound"`
+  -- the observed lower bound of that message's output tokens, not
+  guaranteed to be >= the true count. The group's input-side facts stay
+  `"ok"`, and the token amount itself is unchanged (still priced and
+  included in rows/totals); the flag is a warning only.
   When Anthropic's prompt-cache TTL breakdown (5-minute vs 1-hour writes) is
   present in the log, it's used; otherwise the cache-write tokens are priced
   at the 5-minute rate as an explicit **lower bound** and flagged
@@ -341,7 +350,7 @@ agent-cost measure --session-id <id> [--session-id <id> ...] \
 ```json
 {
   "protocol_version": "measure/v1",
-  "producer_version": "0.2.1",
+  "producer_version": "0.3.0",
   "accounting_basis": "agent-cost-raw-total/v2",
   "generated_at": "...",
   "window": { "since": "...", "until": null },
@@ -362,7 +371,7 @@ agent-cost measure --session-id <id> [--session-id <id> ...] \
     "duplicate_rows_skipped": 0,
     "conflicting_duplicate_groups": 0,
     "missing_dedup_identity_rows": 0,
-    "source_quality": { "ok": 41, "first_event_delta": 2, "identity_missing": 0 }
+    "source_quality": { "ok": 41, "first_event_delta": 2, "identity_missing": 0, "output_lower_bound": 1 }
   }
 }
 ```
@@ -394,8 +403,9 @@ includes absolute file paths, rollout paths, prompt/message content, or git
 branch names -- only the fields needed to reproduce a cost estimate:
 `occurred_at_utc`, `agent`, `session_id`, `model_raw`, `model_key`,
 `token_kind`, `tokens`, `mode`, and `source_quality` (a fixed-vocabulary
-caveat about how that one fact was derived, e.g. `"ok"` or Codex's
-`"first_event_delta"` -- never null).
+caveat about how that one fact was derived: `"ok"`, Codex's
+`"first_event_delta"`, or Claude's `"identity_missing"` /
+`"output_lower_bound"` -- never null).
 
 ## License
 

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/coding_agent_cost.egg-info/SOURCES.txt

@@ -29,4 +29,6 @@ tests/test_rates.py
 tests/test_reader_claude.py
 tests/test_reader_codex.py
 tests/test_reader_dedup.py
+tests/test_reader_output_lower_bound.py
+tests/test_sonnet_5_5_rates.py
 tests/test_version_consistency.py

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "coding-agent-cost"
-version = "0.2.1"
+version = "0.3.0"
 description = "Estimate AI coding agent (Claude Code / Codex CLI) token usage and cost from local logs"
 readme = "README.md"
 license = { text = "MIT" }

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/tests/test_astra_pricing.py

@@ -45,7 +45,7 @@ def test_standard_fast_and_unknown_mode_prices(kind, usd, credits, mode, multipl
 
 def test_catalog_observation_boundary_is_not_backdated():
     catalog = load_rates()
-    assert catalog.catalog_version == "2026-09-23"
+    assert catalog.catalog_version == "2026-09-29"
     entry = catalog.models["gpt-6-astra"]
     assert entry.aliases == ()
     assert entry.rates[0].effective_from == CUTOFF
@@ -149,7 +149,7 @@ def test_measure_reads_isolated_synthetic_astra_with_existing_contract(tmp_path,
     assert cli.main(["measure", "--session-id", "synthetic-astra", "--agent", "codex"]) == 0
     report = json.loads(capsys.readouterr().out)
     assert report["protocol_version"] == "measure/v1"
-    assert report["rates"]["catalog_version"] == "2026-09-23"
+    assert report["rates"]["catalog_version"] == "2026-09-29"
     assert len(report["rates"]["sha256"]) == 64
     assert report["sessions"]["synthetic-astra"]["totals"]["estimated_cost_usd"] == 122
     assert report["total"]["totals"]["credits"] == 3050

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/tests/test_cli.py

@@ -471,6 +471,7 @@ def test_measure_unknown_session_id_exits_zero_with_empty_result(tmp_path, monke
         "ok": 0,
         "first_event_delta": 0,
         "identity_missing": 0,
+        "output_lower_bound": 0,
     }
 
 

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/tests/test_e0a_review3_hardening.py

@@ -19,7 +19,7 @@ architect レビュー (3 巡目) と builder レビューで、既存テスト
        かつ確定金額が計上されない。境界ちょうどは priced になる。
 
 条件D (PR レビュー3点目・追加分): catalog は
-       catalog_version == "2026-09-23"（rates.json の現行値）であり、sources の中に url
+       catalog_version == "2026-09-29"（rates.json の現行値）であり、sources の中に url
        "https://platform.claude.com/docs/en/about-claude/pricing" の
        entry があって、その note に
        "sha256=d79ad28567196bd55dd50e2fd89341b9da9774a45c1d02fcf387078507fd15e0"
@@ -224,7 +224,7 @@ def test_condition_c_claude_fable_5_1_at_effective_from_boundary_is_priced_contr
 
 
 def test_condition_d_catalog_version_and_pricing_source_sha256_note():
-    """条件D: catalog_version == "2026-09-23"（rates.json の現行値）であり、sources に url
+    """条件D: catalog_version == "2026-09-29"（rates.json の現行値）であり、sources に url
     "https://platform.claude.com/docs/en/about-claude/pricing" の entry が
     あって、その note に指定の sha256 文字列を含むこと。
 
@@ -233,7 +233,7 @@ def test_condition_d_catalog_version_and_pricing_source_sha256_note():
     ダイジェストを貼り付けてしまう。
     """
     catalog = load_rates()
-    assert catalog.catalog_version == "2026-09-23"
+    assert catalog.catalog_version == "2026-09-29"
 
     matches = [
         s

coding_agent_cost-0.3.0/tests/test_reader_output_lower_bound.py

@@ -0,0 +1,265 @@
+"""Tests for ``source_quality="output_lower_bound"`` (E0-a4 S1).
+
+Spec: when the row a dedup group adopts (``_select_adopted_row``) has no
+valid ``message.stop_reason`` (a non-empty ``str``), that group's
+``token_kind="output"`` fact is flagged ``"output_lower_bound"`` -- its
+``output_tokens`` is the streaming head's value, a lower bound on the true
+count. Input-side facts from the same group stay ``"ok"``. The decision is
+"is the *adopted* row final", not "does the group contain a final row", so
+a final row followed by a differing non-final row (adoption switches to
+the last row) is flagged too. Identity-missing standalone rows keep
+``"identity_missing"``. Token amounts, adoption and conflict counting are
+unchanged -- only the flag is added.
+
+All fixtures are synthetic and written to ``tmp_path``.
+"""
+
+import json
+
+import pytest
+
+from agent_cost import cli
+from agent_cost.facts import SOURCE_QUALITY_VALUES
+from agent_cost.readers.claude import parse_session_detailed
+
+_ABSENT = object()
+
+
+def _event(
+    ts,
+    output,
+    *,
+    stop_reason=_ABSENT,
+    msg_id="msg-1",
+    req_id="req-1",
+    session_id="s-olb",
+    input_tokens=10,
+    cache_read=20,
+    cache_write_5m=30,
+):
+    message = {
+        "id": msg_id,
+        "model": "claude-sonnet-5",
+        "usage": {
+            "input_tokens": input_tokens,
+            "cache_read_input_tokens": cache_read,
+            "cache_creation_input_tokens": cache_write_5m,
+            "cache_creation": {"ephemeral_5m_input_tokens": cache_write_5m, "ephemeral_1h_input_tokens": 0},
+            "output_tokens": output,
+        },
+    }
+    if msg_id is None:
+        del message["id"]
+    if stop_reason is not _ABSENT:
+        message["stop_reason"] = stop_reason
+    event = {
+        "type": "assistant",
+        "timestamp": ts,
+        "sessionId": session_id,
+        "requestId": req_id,
+        "message": message,
+    }
+    if req_id is None:
+        del event["requestId"]
+    return event
+
+
+def _write(path, events):
+    path.write_text("\n".join(json.dumps(e) for e in events) + "\n")
+    return path
+
+
+def _parse(tmp_path, events):
+    return parse_session_detailed(_write(tmp_path / "session.jsonl", events))
+
+
+def _quality_by_kind(facts):
+    return {f.token_kind: f.source_quality for f in facts}
+
+
+def _tokens_by_kind(facts):
+    return {f.token_kind: f.tokens for f in facts}
+
+
+INPUT_KINDS = ("input_nocache", "cache_read", "cache_write_5m")
+
+
+def test_output_lower_bound_is_last_source_quality_value():
+    assert SOURCE_QUALITY_VALUES[-1] == "output_lower_bound"
+    assert SOURCE_QUALITY_VALUES[:3] == ("ok", "first_event_delta", "identity_missing")
+
+
+def test_group_without_final_row_flags_only_output(tmp_path):
+    """AC1: no row in the group has a stop_reason -> adopted = last row,
+    its output fact is output_lower_bound, input-side facts stay ok."""
+    result = _parse(
+        tmp_path,
+        [
+            _event("2026-10-01T00:00:00Z", 1, stop_reason=None),
+            _event("2026-10-01T00:00:01Z", 1),
+        ],
+    )
+    quality = _quality_by_kind(result.facts)
+    assert quality["output"] == "output_lower_bound"
+    for kind in INPUT_KINDS:
+        assert quality[kind] == "ok"
+    assert _tokens_by_kind(result.facts) == {"input_nocache": 10, "cache_read": 20, "cache_write_5m": 30, "output": 1}
+    assert result.duplicate_rows_skipped == 1
+    assert result.conflicting_duplicate_groups == 0
+
+
+def test_single_row_group_without_final_flags_output(tmp_path):
+    result = _parse(tmp_path, [_event("2026-10-01T00:00:00Z", 4)])
+    quality = _quality_by_kind(result.facts)
+    assert quality["output"] == "output_lower_bound"
+    for kind in INPUT_KINDS:
+        assert quality[kind] == "ok"
+
+
+def test_group_ending_in_final_row_is_all_ok(tmp_path):
+    """AC2: the last row carries a stop_reason -> every fact is ok."""
+    result = _parse(
+        tmp_path,
+        [
+            _event("2026-10-01T00:00:00Z", 1),
+            _event("2026-10-01T00:00:01Z", 2),
+            _event("2026-10-01T00:00:02Z", 57, stop_reason="tool_use"),
+        ],
+    )
+    assert {f.source_quality for f in result.facts} == {"ok"}
+    assert _tokens_by_kind(result.facts)["output"] == 57
+
+
+def test_final_row_followed_by_differing_row_flags_output(tmp_path):
+    """AC2b: a final row followed by a different non-final row switches
+    adoption to the last row (no stop_reason) -> output_lower_bound."""
+    result = _parse(
+        tmp_path,
+        [
+            _event("2026-10-01T00:00:00Z", 5, stop_reason="end_turn"),
+            _event("2026-10-01T00:00:01Z", 8),
+        ],
+    )
+    quality = _quality_by_kind(result.facts)
+    assert quality["output"] == "output_lower_bound"
+    for kind in INPUT_KINDS:
+        assert quality[kind] == "ok"
+    assert _tokens_by_kind(result.facts)["output"] == 8
+
+
+def test_final_row_followed_by_identical_row_stays_ok(tmp_path):
+    """AC2c: a final row followed by an identical (non-final) row keeps the
+    final row adopted -> ok, even though the group's last row isn't final."""
+    result = _parse(
+        tmp_path,
+        [
+            _event("2026-10-01T00:00:00Z", 5, stop_reason="end_turn"),
+            _event("2026-10-01T00:00:01Z", 5),
+        ],
+    )
+    assert {f.source_quality for f in result.facts} == {"ok"}
+    assert _tokens_by_kind(result.facts)["output"] == 5
+
+
+@pytest.mark.parametrize("stop_reason", ["", None, 0, 1, 1.5, False, ["end_turn"], _ABSENT])
+def test_invalid_stop_reason_only_group_flags_output(tmp_path, stop_reason):
+    """AC2c: stop_reason "" / None / numbers / non-str only -> not final."""
+    result = _parse(
+        tmp_path,
+        [
+            _event("2026-10-01T00:00:00Z", 3, stop_reason=stop_reason),
+            _event("2026-10-01T00:00:01Z", 6, stop_reason=stop_reason),
+        ],
+    )
+    quality = _quality_by_kind(result.facts)
+    assert quality["output"] == "output_lower_bound"
+    for kind in INPUT_KINDS:
+        assert quality[kind] == "ok"
+    assert _tokens_by_kind(result.facts)["output"] == 6
+
+
+def test_final_row_with_zero_output_is_ok(tmp_path):
+    """AC2c: a final row with output 0 emits no output fact; the rest is ok."""
+    result = _parse(
+        tmp_path,
+        [
+            _event("2026-10-01T00:00:00Z", 0),
+            _event("2026-10-01T00:00:01Z", 0, stop_reason="end_turn"),
+        ],
+    )
+    assert "output" not in _tokens_by_kind(result.facts)
+    assert {f.source_quality for f in result.facts} == {"ok"}
+
+
+@pytest.mark.parametrize("missing", ["msg_id", "req_id"])
+def test_identity_missing_row_stays_identity_missing(tmp_path, missing):
+    """AC2d: a standalone row without a full id pair is not a group, so it
+    keeps identity_missing on every fact (including output)."""
+    kwargs = {missing: None}
+    result = _parse(tmp_path, [_event("2026-10-01T00:00:00Z", 9, **kwargs)])
+    assert result.missing_dedup_identity_rows == 1
+    assert "output" in _tokens_by_kind(result.facts)
+    assert {f.source_quality for f in result.facts} == {"identity_missing"}
+
+
+def test_flag_is_per_group(tmp_path):
+    """One group without a final row, one with: only the former's output is flagged."""
+    result = _parse(
+        tmp_path,
+        [
+            _event("2026-10-01T00:00:00Z", 2, msg_id="msg-a", req_id="req-a"),
+            _event("2026-10-01T00:00:01Z", 40, msg_id="msg-b", req_id="req-b", stop_reason="end_turn"),
+        ],
+    )
+    output_facts = [f for f in result.facts if f.token_kind == "output"]
+    assert [(f.tokens, f.source_quality) for f in output_facts] == [(2, "output_lower_bound"), (40, "ok")]
+    assert all(f.source_quality == "ok" for f in result.facts if f.token_kind != "output")
+
+
+# ---------------------------------------------------------------------------
+# CLI (AC3): measure's data_quality.source_quality always carries the key.
+# ---------------------------------------------------------------------------
+
+
+def _setup_claude_session(tmp_path, monkeypatch, events):
+    claude_home = tmp_path / "claude_home"
+    codex_home = tmp_path / "codex_home"
+    project_dir = claude_home / "projects" / "olb"
+    project_dir.mkdir(parents=True)
+    codex_home.mkdir()
+    monkeypatch.setenv("CLAUDE_HOME", str(claude_home))
+    monkeypatch.setenv("CODEX_HOME", str(codex_home))
+    monkeypatch.delenv("AGENT_COST_CONFIG", raising=False)
+    _write(project_dir / "session.jsonl", events)
+
+
+def _measure_source_quality(capsys):
+    rc = cli.main(["measure", "--session-id", "s-olb", "--since", "2026-09-01"])
+    assert rc == 0
+    return json.loads(capsys.readouterr().out)["data_quality"]["source_quality"]
+
+
+def test_measure_counts_output_lower_bound(tmp_path, monkeypatch, capsys):
+    _setup_claude_session(
+        tmp_path,
+        monkeypatch,
+        [
+            _event("2026-10-01T00:00:00Z", 1, msg_id="msg-a", req_id="req-a"),
+            _event("2026-10-01T00:00:01Z", 30, msg_id="msg-b", req_id="req-b", stop_reason="end_turn"),
+        ],
+    )
+    sq = _measure_source_quality(capsys)
+    assert set(sq.keys()) == set(SOURCE_QUALITY_VALUES)
+    assert sq["output_lower_bound"] == 1
+    assert sq["ok"] == 7
+    assert sq["identity_missing"] == 0
+
+
+def test_measure_reports_zero_output_lower_bound_when_absent(tmp_path, monkeypatch, capsys):
+    _setup_claude_session(
+        tmp_path,
+        monkeypatch,
+        [_event("2026-10-01T00:00:00Z", 30, stop_reason="end_turn")],
+    )
+    sq = _measure_source_quality(capsys)
+    assert sq == {"ok": 4, "first_event_delta": 0, "identity_missing": 0, "output_lower_bound": 0}

coding_agent_cost-0.3.0/tests/test_sonnet_5_5_rates.py

@@ -0,0 +1,69 @@
+"""Acceptance tests for the claude-sonnet-5-5 rate entry (catalog_version 2026-09-29).
+
+Primary sources: platform.claude.com/docs/en/about-claude/pricing (Sonnet 5.5
+row, no cache_read footnote), anthropic.com/claude-sonnet-5-5 (launch
+2026-09-28), Claude Code CHANGELOG 2.1.284 (Sonnet 5.5 is the default Sonnet).
+"""
+
+from datetime import datetime, timezone
+from decimal import Decimal
+
+from agent_cost.rates import load_rates
+
+UTC = timezone.utc
+
+
+def dt(s):
+    return datetime.fromisoformat(s).replace(tzinfo=UTC)
+
+
+def test_claude_sonnet_5_5_resolves_with_pricing_page_values_on_or_after_launch():
+    catalog = load_rates()
+    for occurred_at in ("2026-09-28T00:00:00+00:00", "2026-12-01T00:00:00+00:00"):
+        resolved, period = catalog.rate_for("claude-sonnet-5-5", dt(occurred_at))
+        assert resolved == "claude-sonnet-5-5"
+        assert period is not None
+        assert period.rate_id == "claude-sonnet-5-5-launch-2026-09-28"
+        assert period.values["input_nocache"] == Decimal("2.0")
+        assert period.values["cache_read"] == Decimal("0.20")
+        assert period.values["cache_write_5m"] == Decimal("2.50")
+        assert period.values["cache_write_1h"] == Decimal("4.0")
+        assert period.values["output"] == Decimal("10.0")
+
+
+def test_claude_sonnet_5_5_is_unpriced_before_launch_date():
+    catalog = load_rates()
+    resolved, before = catalog.rate_for("claude-sonnet-5-5", dt("2026-09-27T23:59:59+00:00"))
+    assert resolved == "claude-sonnet-5-5"
+    assert before is None
+
+
+def test_claude_sonnet_5_5_is_its_own_entry_not_an_alias_of_sonnet_5():
+    # Same values as claude-sonnet-5 today, but the pricing page lists the rows
+    # independently; a future change to one must not silently apply to the other.
+    catalog = load_rates()
+    assert "claude-sonnet-5-5" in catalog.models
+    assert catalog.models["claude-sonnet-5-5"].aliases == ()
+    assert "claude-sonnet-5-5" not in catalog.models["claude-sonnet-5"].aliases
+    _, s5 = catalog.rate_for("claude-sonnet-5", dt("2026-09-28T00:00:00+00:00"))
+    _, s55 = catalog.rate_for("claude-sonnet-5-5", dt("2026-09-28T00:00:00+00:00"))
+    assert s5.rate_id != s55.rate_id
+    assert s5.values == s55.values
+
+
+def test_claude_sonnet_5_5_cache_read_is_standard_multiplier_and_no_fast_mode():
+    catalog = load_rates()
+    entry = catalog.models["claude-sonnet-5-5"]
+    assert entry.fast_multiplier == Decimal("1.0")
+    period = entry.rates[0]
+    assert period.values["cache_read"] == period.values["input_nocache"] * Decimal("0.1")
+    assert period.values["cache_write_5m"] == period.values["input_nocache"] * Decimal("1.25")
+    assert period.values["cache_write_1h"] == period.values["input_nocache"] * Decimal("2")
+
+
+def test_catalog_version_and_sources_record_the_sonnet_5_5_primary_sources():
+    catalog = load_rates()
+    assert catalog.catalog_version == "2026-09-29"
+    urls = {(s.get("url"), s.get("retrieved_at")) for s in catalog.sources}
+    assert ("https://platform.claude.com/docs/en/about-claude/pricing", "2026-09-29") in urls
+    assert ("https://www.anthropic.com/claude-sonnet-5-5", "2026-09-29") in urls

{coding_agent_cost-0.2.1 → coding_agent_cost-0.3.0}/tests/test_version_consistency.py

@@ -1,8 +1,8 @@
 """Spec: pyproject.toml's [project].version and agent_cost.__version__ must
-agree, and both must read "0.2.1" for this change (the Opus 5.5 catalog release bumps the package
-version alongside the measure/v1 data_quality additions). A broken
-implementation would bump one file but not the other, or forget the bump
-entirely.
+agree, and both must read "0.3.0" for this change (the output_lower_bound
+source_quality value is an additive vocabulary change, so it bumps the minor
+version). A broken implementation would bump one file but not the other, or
+forget the bump entirely.
 
 Uses a plain regex instead of tomllib/tomli: this repo is dependency-free
 by design (pyproject.toml's own dependencies = []) and tomllib is
@@ -29,5 +29,5 @@ def test_pyproject_version_matches_package_version():
 
 
 def test_version_is_0_2_0():
-    assert agent_cost.__version__ == "0.2.1"
-    assert _pyproject_version() == "0.2.1"
+    assert agent_cost.__version__ == "0.3.0"
+    assert _pyproject_version() == "0.3.0"