rote-cli 0.15.0__tar.gz → 0.16.0__tar.gz

{rote_cli-0.15.0 → rote_cli-0.16.0}/LICENSE

@@ -1,3 +1,4 @@
+
                                  Apache License
                            Version 2.0, January 2004
                         http://www.apache.org/licenses/
@@ -137,8 +138,8 @@
 
    6. Trademarks. This License does not grant permission to use the trade
       names, trademarks, service marks, or product names of the Licensor,
-      except as required for describing the origin of the Work and
-      reproducing the content of the NOTICE file.
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
 
    7. Disclaimer of Warranty. Unless required by applicable law or
       agreed to in writing, Licensor provides the Work (and each
@@ -163,15 +164,15 @@
       has been advised of the possibility of such damages.
 
    9. Accepting Warranty or Additional Liability. While redistributing
-      the Work or Derivative Works thereof, You may accept and charge a
-      fee for acceptance of support, warranty, indemnity, or other liability
-      obligations and/or rights consistent with this License. However, in
-      accepting such obligations, You may act only on Your own behalf and on
-      Your sole responsibility, not on behalf of any other Contributor, and
-      only if You agree to indemnify, defend, and hold each Contributor
-      harmless for any liability incurred by, or claims asserted against,
-      such Contributor by reason of your accepting any such warranty or
-      additional liability.
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
 
    END OF TERMS AND CONDITIONS
 

{rote_cli-0.15.0 → rote_cli-0.16.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: rote-cli
-Version: 0.15.0
+Version: 0.16.0
 Summary: Compile AI agent skills into cheap, fast, deterministic pipelines that run without an LLM in the loop
 Project-URL: Homepage, https://roteskills.com
 Project-URL: Repository, https://github.com/trevhud/rote
@@ -24,6 +24,7 @@ Classifier: Topic :: Software Development :: Code Generators
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Classifier: Typing :: Typed
 Requires-Python: >=3.11
+Requires-Dist: graphql-core<4,>=3.3
 Requires-Dist: pydantic>=2.7
 Requires-Dist: pytest<10,>=9.0
 Requires-Dist: pyyaml>=6.0
@@ -35,10 +36,15 @@ Provides-Extra: dev
 Requires-Dist: anthropic>=1.15; extra == 'dev'
 Requires-Dist: dbos<4,>=3.3; extra == 'dev'
 Requires-Dist: fastmcp<5,>=4.1.1; extra == 'dev'
+Requires-Dist: google-api-python-client>=2.201.0; extra == 'dev'
+Requires-Dist: google-auth-httplib2>=0.4.4; extra == 'dev'
+Requires-Dist: google-auth>=2.62.0; extra == 'dev'
+Requires-Dist: httplib2>=0.32.0; extra == 'dev'
 Requires-Dist: httpx>=0.28; extra == 'dev'
 Requires-Dist: mypy>=1.10; extra == 'dev'
 Requires-Dist: openai>=2.45; extra == 'dev'
 Requires-Dist: pytest-asyncio>=0.23; extra == 'dev'
+Requires-Dist: requests>=2.34.2; extra == 'dev'
 Requires-Dist: ruff>=0.6; extra == 'dev'
 Requires-Dist: temporalio>=1.7; extra == 'dev'
 Requires-Dist: trustme>=1.2.1; extra == 'dev'
@@ -238,13 +244,27 @@ and want to run a thousand more, unattended.
 
 ## How it works
 
-One `rote compile` run does the whole thing. An LLM agent (itself
-defined as a skill) reads the source skill, applies a structured
-compilation rubric, and emits a runtime-agnostic intermediate
-representation (`pipeline.yaml`), extracted Python modules for the
-deterministic parts, typed signature stubs for the LLM-judge parts, and
+One `rote compile` run does the whole thing. First rote decides the
+pipeline's structure: which nodes exist and of which kind, how they are
+wired, which run once per item, which are mandatory, where a person must
+approve, and which constants they carry (System-1, see
+[`docs/design/system1-compile.md`](docs/design/system1-compile.md)). One
+Claude Haiku call reads the whole skill and proposes that structure, and
+rote checks every node against the skill's text, its tools' schemas and
+the sample input before accepting it. `--structure rules` decides it with
+rules over the skill's text instead, with no model call.
+Then an LLM agent (itself defined as a skill) writes only the text those
+decisions need: the extracted Python modules for the deterministic
+parts, their tests, and the typed signatures and prompts for the
+LLM-judge parts. rote holds the agent to its decisions, and emits a
+runtime-agnostic intermediate representation (`pipeline.yaml`) plus
 runnable code for the durable execution engine of your choice.
 
+`--no-system1` runs the original path instead, where the agent applies
+the full compilation rubric and decides the structure too. `--update`
+always uses that path, because System-1 does not support incremental
+re-compilation yet.
+
 The agent implements the deterministic `pure_function` steps the skill
 spells out and writes tests for them. Before the compile reports
 success, rote checks what the agent wrote and runs those tests, and any
@@ -443,7 +463,7 @@ through your API credentials.
 - **`rote config`**: print every configurable default with its
   effective value *and the layer that set it*. Resolution everywhere is
   `flag > ROTE_* env (ROTE_RUNTIME, ROTE_DEPLOY, ROTE_AGENT,
-  ROTE_MODEL) > project rote.yaml > user config > built-in`. Config
+  ROTE_MODEL, ROTE_EFFORT) > project rote.yaml > user config > built-in`. Config
   files are strict: a typo'd key or value is a loud error, never a
   silent fallback. `--json` for automation.
 - **`rote eval <compiled>`**: render the before/after scorecard (wall
@@ -484,6 +504,11 @@ way, prefer the more deterministic kind.
 Pick with `--runtime`; the same IR drives all of them. Under
 `--backend api`, none of the emitted code references MCP: the
 crystallization step replaces tool calls with direct vendor API calls.
+The compiler agent gets web search and fetch tools for that, writes each
+call from the vendor's current reference page, and cites the page as
+`# docs: <url>`; `compile-report.md` lists every cited page beside
+whether the agent fetched it during the compile. The api backend's calls
+are Python, so it needs a Python runtime (`dbos`, `python`, `temporal`).
 Under the default `--backend mcp`, tool-using nodes emit a working MCP
 client call (with durable park-on-auth on every MCP-capable runtime);
 see [`docs/mcp-client.md`](docs/mcp-client.md).

{rote_cli-0.15.0 → rote_cli-0.16.0}/README.md

@@ -181,13 +181,27 @@ and want to run a thousand more, unattended.
 
 ## How it works
 
-One `rote compile` run does the whole thing. An LLM agent (itself
-defined as a skill) reads the source skill, applies a structured
-compilation rubric, and emits a runtime-agnostic intermediate
-representation (`pipeline.yaml`), extracted Python modules for the
-deterministic parts, typed signature stubs for the LLM-judge parts, and
+One `rote compile` run does the whole thing. First rote decides the
+pipeline's structure: which nodes exist and of which kind, how they are
+wired, which run once per item, which are mandatory, where a person must
+approve, and which constants they carry (System-1, see
+[`docs/design/system1-compile.md`](docs/design/system1-compile.md)). One
+Claude Haiku call reads the whole skill and proposes that structure, and
+rote checks every node against the skill's text, its tools' schemas and
+the sample input before accepting it. `--structure rules` decides it with
+rules over the skill's text instead, with no model call.
+Then an LLM agent (itself defined as a skill) writes only the text those
+decisions need: the extracted Python modules for the deterministic
+parts, their tests, and the typed signatures and prompts for the
+LLM-judge parts. rote holds the agent to its decisions, and emits a
+runtime-agnostic intermediate representation (`pipeline.yaml`) plus
 runnable code for the durable execution engine of your choice.
 
+`--no-system1` runs the original path instead, where the agent applies
+the full compilation rubric and decides the structure too. `--update`
+always uses that path, because System-1 does not support incremental
+re-compilation yet.
+
 The agent implements the deterministic `pure_function` steps the skill
 spells out and writes tests for them. Before the compile reports
 success, rote checks what the agent wrote and runs those tests, and any
@@ -386,7 +400,7 @@ through your API credentials.
 - **`rote config`**: print every configurable default with its
   effective value *and the layer that set it*. Resolution everywhere is
   `flag > ROTE_* env (ROTE_RUNTIME, ROTE_DEPLOY, ROTE_AGENT,
-  ROTE_MODEL) > project rote.yaml > user config > built-in`. Config
+  ROTE_MODEL, ROTE_EFFORT) > project rote.yaml > user config > built-in`. Config
   files are strict: a typo'd key or value is a loud error, never a
   silent fallback. `--json` for automation.
 - **`rote eval <compiled>`**: render the before/after scorecard (wall
@@ -427,6 +441,11 @@ way, prefer the more deterministic kind.
 Pick with `--runtime`; the same IR drives all of them. Under
 `--backend api`, none of the emitted code references MCP: the
 crystallization step replaces tool calls with direct vendor API calls.
+The compiler agent gets web search and fetch tools for that, writes each
+call from the vendor's current reference page, and cites the page as
+`# docs: <url>`; `compile-report.md` lists every cited page beside
+whether the agent fetched it during the compile. The api backend's calls
+are Python, so it needs a Python runtime (`dbos`, `python`, `temporal`).
 Under the default `--backend mcp`, tool-using nodes emit a working MCP
 client call (with durable park-on-auth on every MCP-capable runtime);
 see [`docs/mcp-client.md`](docs/mcp-client.md).

{rote_cli-0.15.0 → rote_cli-0.16.0}/examples/expense-audit/README.md

@@ -104,6 +104,7 @@ examples/expense-audit/
 ├── scorecard.md                  # measured before/after (rote eval --run)
 ├── skill/
 │   ├── SKILL.md                  # the source skill, as an agent would run it
+│   ├── sample-input.json         # the sample run `rote compile` grounds its contracts in
 │   └── references/
 │       ├── expense-policy.md     # caps, receipts, severities
 │       ├── fx-rates.md           # finance month-end rates

rote_cli-0.16.0/examples/expense-audit/fidelity/cases.yaml

@@ -0,0 +1,68 @@
+# Behavioral fidelity cases for `rote bench` (docs/compile-bench.md).
+#
+# The skill runs as an agent on each case (the committed reference.json),
+# and each compiled pipeline runs on the same cases. Every case has more
+# than one employee and more than one finding, so a fan-out or a check that
+# silently does nothing cannot pass. The controller's decisions are fixed.
+
+output_fields:
+  period: the reporting period, YYYY-MM
+  batch_total_usd: the whole batch total in USD, a number
+  auto_approved_employees: list of the names of the employees whose reports were auto-approved
+  duplicate_ids: list of the expense ids flagged as duplicates
+  late_ids: list of the expense ids flagged as late
+  blocked_ids: list of the expense ids with at least one block finding
+  review_queue: >-
+    the review queue, a list with one object per finding, each with
+    `expense_id`, `employee` and `severity` (block or review)
+  reimbursable_usd: the reimbursable total in USD after the controller's decisions, a number
+  returned_usd: the total returned to employees in USD, a number
+  pending_usd: the total still pending a decision in USD, a number
+  report_status: the status the filed report is marked with (final or draft)
+  report_filename: the file name the audit report is saved as
+
+cases:
+  - id: base
+    description: the bundled September batch; the controller approves three review items and rejects one
+    input_file: ../input.json
+    gates:
+      - reviewer: Priya Shah
+        approved_ids: [EXP-2403, EXP-2413, EXP-2420]
+        rejected_ids: [EXP-2410]
+
+  - id: second_batch
+    description: >-
+      the same batch with a missing receipt (block), a line dated after the
+      period closed (late), a category the policy does not list (block), and
+      an entertainment line that now carries its pre-approval; the controller
+      approves two review items and rejects two
+    input_file: ../input.json
+    input_patches:
+      - {set: submissions.13.receipt, value: false}
+      - {set: submissions.14.date, value: "2026-10-02"}
+      - {set: submissions.8.preapproval_id, value: PA-118}
+      - append: submissions
+        value:
+          id: EXP-2442
+          employee: Emi Tanaka
+          department: Design
+          date: "2026-09-27"
+          category: gifts
+          merchant: Tokyu Hands
+          amount: 4500
+          currency: JPY
+          receipt: true
+    gates:
+      - reviewer: Priya Shah
+        approved_ids: [EXP-2410, EXP-2440]
+        rejected_ids: [EXP-2403, EXP-2413]
+
+  - id: unknown_currency
+    description: a line in a currency the rate table does not list, so the skill must stop before auditing
+    input_file: ../input.json
+    input_patches:
+      - {set: submissions.10.currency, value: CHF}
+    gates:
+      - reviewer: Priya Shah
+        approved_ids: [EXP-2403, EXP-2413, EXP-2420]
+        rejected_ids: [EXP-2410]