feat(m-decomp): upgraded pipeline and added README, examples, and fixed module issues (#676)

* upd: clean over long or result files

* add: validation code generation

* add: validation code icl

* add: prompt init

* add: validation decision

* add: decomp jinja

* refact: pipeline and primary stages

* refact: module logging

* fea: validation report

* fea: validation report icl

* fea: cli script with config

* add: examples

* add: examples

* add: README doc

* pre-commit: add test attribute

* upd: type annotations

* fix: add constraint type annotation

* upd: pre-commit format

* upd: pre-commit type annotations

* upd: pre-commit format

* add: m_decompose tests

* add: constraint retry

* upd: constraint

* upd: constraint

* upd: same logmode

* fix: a missed parse

* pre-commit format

* add: multi request support

* clean

* fix: type clean

* upd: input file arg

* fea: constraint retry

* test

* fix

* fix

* fix

* clean: final result

* add: decompose tests

* add: decompose tests

* clean: pre-commit format

* clean: pre-commit formating on decompose tests

Author: csbobby

cli/decompose/decompose.py

@@ -0,0 +1,44 @@
+import logging
+import sys
+from enum import StrEnum
+
+
+class LogMode(StrEnum):
+    demo = "demo"
+    debug = "debug"
+
+
+_CONFIGURED = False
+
+
+def configure_logging(log_mode: LogMode = LogMode.demo) -> None:
+    global _CONFIGURED
+
+    level = logging.DEBUG if log_mode == LogMode.debug else logging.INFO
+
+    root_logger = logging.getLogger()
+
+    if not _CONFIGURED:
+        handler = logging.StreamHandler(sys.stdout)
+        handler.setFormatter(
+            logging.Formatter("[%(levelname)s] %(name)s | %(message)s")
+        )
+        root_logger.handlers.clear()
+        root_logger.addHandler(handler)
+        _CONFIGURED = True
+
+    root_logger.setLevel(level)
+
+    logging.getLogger("m_decompose").setLevel(level)
+    logging.getLogger("mellea").setLevel(level)
+
+
+def get_logger(name: str) -> logging.Logger:
+    return logging.getLogger(name)
+
+
+def log_section(logger: logging.Logger, title: str) -> None:
+    logger.info("")
+    logger.info("=" * 72)
+    logger.info(title)
+    logger.info("=" * 72)

cli/decompose/logging.py

@@ -4,6 +4,14 @@ import os
 import textwrap
 
 import mellea
+{%- if "code" in identified_constraints | map(attribute="val_strategy") %}
+from mellea.stdlib.requirement import req
+{% for c in identified_constraints %}
+{%- if c.val_fn %}
+from validations.{{ c.val_fn_name }} import validate_input as {{ c.val_fn_name }}
+{%- endif %}
+{%- endfor %}
+{%- endif %}
 
 m = mellea.start_session()
 {%- if user_inputs %}
@@ -30,7 +38,14 @@ except KeyError as e:
     {%- if item.constraints %}
     requirements=[
         {%- for c in item.constraints %}
+        {%- if c.val_fn %}
+        req(
+            {{ c.constraint | tojson}},
+            validation_fn={{ c.val_fn_name }},
+        ),
+        {%- else %}
         {{ c.constraint | tojson}},
+        {%- endif %}
         {%- endfor %}
     ],
     {%- else %}

cli/decompose/m_decomp_result_v1.py.jinja2

@@ -0,0 +1,91 @@
+{% if user_inputs -%}
+import os
+{% endif -%}
+import textwrap
+
+import mellea
+{%- if "code" in identified_constraints | map(attribute="val_strategy") %}
+from mellea.stdlib.requirement import req
+{% for c in identified_constraints %}
+{%- if c.val_fn %}
+from validations.{{ c.val_fn_name }} import validate_input as {{ c.val_fn_name }}
+{%- endif %}
+{%- endfor %}
+{%- endif %}
+
+m = mellea.start_session()
+{%- if user_inputs %}
+
+
+# User Input Variables
+try:
+    {%- for var in user_inputs %}
+    {{ var | lower }} = os.environ["{{ var | upper }}"]
+    {%- endfor %}
+except KeyError as e:
+    print(f"ERROR: One or more required environment variables are not set; {e}")
+    exit(1)
+{%- endif %}
+{%- for item in subtasks %}
+
+
+{{ item.tag | lower }}_gnrl = textwrap.dedent(
+    R"""
+    {{ item.general_instructions | trim | indent(width=4, first=False) }}
+    """.strip()
+)
+{{ item.tag | lower }} = m.instruct(
+    {%- if not item.input_vars_required %}
+    {{ item.subtask[3:] | trim | tojson }},
+    {%- else %}
+    textwrap.dedent(
+        R"""
+        {{ item.subtask[3:] | trim }}
+
+        Here are the input variables and their content:
+        {%- for var in item.input_vars_required %}
+
+        - {{ var | upper }} = {{ "{{" }}{{ var | upper }}{{ "}}" }}
+        {%- endfor %}
+        """.strip()
+    ),
+    {%- endif %}
+    {%- if item.constraints %}
+    requirements=[
+        {%- for c in item.constraints %}
+        {%- if c.val_fn %}
+        req(
+            {{ c.constraint | tojson}},
+            validation_fn={{ c.val_fn_name }},
+        ),
+        {%- else %}
+        {{ c.constraint | tojson}},
+        {%- endif %}
+        {%- endfor %}
+    ],
+    {%- else %}
+    requirements=None,
+    {%- endif %}
+    {%- if item.input_vars_required %}
+    user_variables={
+        {%- for var in item.input_vars_required %}
+        {{ var | upper | tojson }}: {{ var | lower }},
+        {%- endfor %}
+    },
+    {%- endif %}
+    grounding_context={
+        "GENERAL_INSTRUCTIONS": {{ item.tag | lower }}_gnrl,
+        {%- for var in item.depends_on %}
+        {{ var | upper | tojson }}: {{ var | lower }}.value,
+        {%- endfor %}
+    },
+)
+assert {{ item.tag | lower }}.value is not None, 'ERROR: task "{{ item.tag | lower }}" execution failed'
+{%- if loop.last %}
+
+
+final_answer = {{ item.tag | lower }}.value
+
+print(final_answer)
+{%- endif -%}
+{%- endfor -%}