Add Lliam UoW (#177)

* Add UoW and adapters to Lliam
* Update Lliam repository to allow injectable FS
* Make batch num padding dynamic, and clean up a few things.

Author: cpyle0819

aws_doc_sdk_examples_tools/lliam/adapters/repository.py

@@ -0,0 +1,132 @@
+from itertools import islice
+from math import ceil, floor, log
+from pathlib import Path
+from typing import Any, Dict, Generator, Iterable, List, Tuple
+
+from aws_doc_sdk_examples_tools.doc_gen import DocGen, Example
+from aws_doc_sdk_examples_tools.fs import Fs, PathFs
+from aws_doc_sdk_examples_tools.lliam.domain.model import Prompt
+from aws_doc_sdk_examples_tools.lliam.config import BATCH_PREFIX
+
+DEFAULT_METADATA_PREFIX = "DEFAULT"
+DEFAULT_BATCH_SIZE = 150
+IAM_POLICY_LANGUAGE = "IAMPolicyGrammar"
+
+
+def batched(iterable: Iterable, n: int) -> Generator[Tuple, Any, None]:
+    "Batch data into tuples of length n. The last batch may be shorter."
+    # batched('ABCDEFG', 3) --> ABC DEF G
+    if n < 1:
+        raise ValueError("n must be at least one")
+    it = iter(iterable)
+    while batch := tuple(islice(it, n)):
+        yield batch
+
+
+class PromptRepository:
+    to_write: Dict[str, str] = {}
+
+    def __init__(self, fs: Fs = PathFs()):
+        self.fs = fs
+
+    def rollback(self):
+        # TODO: This is not what rollback is for. We should be rolling back any
+        # file changes
+        self.to_write = {}
+
+    def add(self, prompt: Prompt):
+        self.to_write[prompt.id] = prompt.content
+
+    def all_all(self, prompts: Iterable[Prompt]):
+        for prompt in prompts:
+            self.add(prompt)
+
+    def batch(self, prompts: Iterable[Prompt]):
+        prompt_list = list(prompts)
+
+        if not prompt_list:
+            return
+
+        batches_count = ceil(len(prompt_list) / DEFAULT_BATCH_SIZE)
+        padding = floor(log(batches_count, 10)) + 1
+        for batch_num, batch in enumerate(batched(prompts, DEFAULT_BATCH_SIZE)):
+            batch_name = f"{BATCH_PREFIX}{(batch_num + 1):0{padding}}"
+            for prompt in batch:
+                prompt.id = f"{batch_name}/{prompt.id}"
+                self.add(prompt)
+
+    def commit(self):
+        base_path = Path(self.partition) if self.partition else Path(".")
+
+        for id, content in self.to_write.items():
+            if content:
+                full_path = base_path / id
+                self.fs.mkdir(full_path.parent)
+                self.fs.write(full_path, content)
+
+    def get(self, id: str):
+        return Prompt(id, self.fs.read(Path(id)))
+
+    def get_all(self, ids: List[str]) -> List[Prompt]:
+        prompts = []
+        for id in ids:
+            prompt = self.get(id)
+            prompts.append(prompt)
+        return prompts
+
+    def set_partition(self, name: str):
+        self.partition_name = name
+
+    @property
+    def partition(self):
+        return self.partition_name or ""
+
+
+class DocGenRepository:
+    def __init__(self, fs: Fs = PathFs()):
+        self.fs = fs
+
+    def rollback(self):
+        # TODO: This is not what rollback is for. We should be rolling back any
+        # file changes
+        self._doc_gen = None
+
+    def get_new_prompts(self, doc_gen_root: str) -> List[Prompt]:
+        # Right now this is the only instance of DocGen used in this Repository,
+        # but if that changes we need to move it up.
+        self._doc_gen = DocGen.from_root(Path(doc_gen_root), fs=self.fs)
+        self._doc_gen.collect_snippets()
+        new_examples = self._get_new_examples()
+        prompts = self._examples_to_prompts(new_examples)
+        return prompts
+
+    def _get_new_examples(self) -> List[Tuple[str, Example]]:
+        examples = self._doc_gen.examples
+
+        filtered_examples: List[Tuple[str, Example]] = []
+        for example_id, example in examples.items():
+            # TCXContentAnalyzer prefixes new metadata title/title_abbrev entries with
+            # the DEFAULT_METADATA_PREFIX. Checking this here to make sure we're only
+            # running the LLM tool on new extractions.
+            title = example.title or ""
+            title_abbrev = example.title_abbrev or ""
+            if title.startswith(DEFAULT_METADATA_PREFIX) and title_abbrev.startswith(
+                DEFAULT_METADATA_PREFIX
+            ):
+                filtered_examples.append((example_id, example))
+        return filtered_examples
+
+    def _examples_to_prompts(self, examples: List[Tuple[str, Example]]) -> List[Prompt]:
+        snippets = self._doc_gen.snippets
+        prompts = []
+        for example_id, example in examples:
+            key = (
+                example.languages[IAM_POLICY_LANGUAGE]
+                .versions[0]
+                .excerpts[0]
+                .snippet_files[0]
+                .replace("/", ".")
+            )
+            snippet = snippets.get(key)
+            prompts.append(Prompt(f"{example_id}.md", snippet.code))
+        return prompts

aws_doc_sdk_examples_tools/lliam/config.py

@@ -0,0 +1,5 @@
+from pathlib import Path
+
+AILLY_DIR = ".ailly_iam_policy"
+AILLY_DIR_PATH = Path(AILLY_DIR)
+BATCH_PREFIX = "batch_"

aws_doc_sdk_examples_tools/lliam/domain/commands.py

@@ -9,7 +9,7 @@ class Command:
 
 @dataclass
 class CreatePrompts(Command):
-    doc_gen_root: str 
+    doc_gen_root: str
     system_prompts: List[str]
     out_dir: str
 

aws_doc_sdk_examples_tools/lliam/service_layer/__init__.py

@@ -0,0 +1,18 @@
+import logging
+
+from aws_doc_sdk_examples_tools.lliam.domain.operations import build_ailly_config
+from aws_doc_sdk_examples_tools.lliam.domain.commands import CreatePrompts
+from aws_doc_sdk_examples_tools.lliam.service_layer.unit_of_work import FsUnitOfWork
+
+logger = logging.getLogger(__name__)
+
+
+def create_prompts(cmd: CreatePrompts, uow: FsUnitOfWork):
+    with uow:
+        system_prompts = uow.prompts.get_all(cmd.system_prompts)
+        ailly_config = build_ailly_config(system_prompts)
+        prompts = uow.doc_gen.get_new_prompts(cmd.doc_gen_root)
+        uow.prompts.batch(prompts)
+        uow.prompts.add(ailly_config)
+        uow.prompts.set_partition(cmd.out_dir)
+        uow.commit()

aws_doc_sdk_examples_tools/lliam/service_layer/create_prompts.py

@@ -0,0 +1,24 @@
+from aws_doc_sdk_examples_tools.fs import Fs, PathFs
+from aws_doc_sdk_examples_tools.lliam.adapters.repository import (
+    PromptRepository,
+    DocGenRepository,
+)
+
+
+class FsUnitOfWork:
+    def __init__(self, fs: Fs = PathFs()):
+        self.fs = fs
+
+    def __enter__(self):
+        self.prompts = PromptRepository(fs=self.fs)
+        self.doc_gen = DocGenRepository(fs=self.fs)
+
+    def __exit__(self, *args):
+        self.rollback()
+
+    def commit(self):
+        self.prompts.commit()
+
+    def rollback(self):
+        self.prompts.rollback()
+        self.doc_gen.rollback()

aws_doc_sdk_examples_tools/lliam/service_layer/unit_of_work.py

@@ -0,0 +1,31 @@
+from pathlib import Path
+
+from aws_doc_sdk_examples_tools.fs import RecordFs
+from aws_doc_sdk_examples_tools.lliam.adapters.repository import PromptRepository
+from aws_doc_sdk_examples_tools.lliam.domain.model import Prompt
+
+
+def test_batch_naming_occurs_properly():
+    """Test that batch naming occurs properly when batching prompts."""
+    fs = RecordFs({})
+    repo = PromptRepository(fs=fs)
+
+    prompts = []
+    for i in range(300):
+        prompts.append(Prompt(f"prompt_{i}.md", f"Content for prompt {i}"))
+
+    repo.batch(prompts)
+
+    expected_batch_1_prompts = 150
+    expected_batch_2_prompts = 150
+
+    batch_1_count = 0
+    batch_2_count = 0
+    for prompt_id in repo.to_write:
+        if prompt_id.startswith("batch_1/"):
+            batch_1_count += 1
+        elif prompt_id.startswith("batch_2/"):
+            batch_2_count += 1
+
+    assert batch_1_count == expected_batch_1_prompts
+    assert batch_2_count == expected_batch_2_prompts

aws_doc_sdk_examples_tools/lliam/test/repository_test.py

@@ -0,0 +1,32 @@
+from pathlib import Path
+
+from aws_doc_sdk_examples_tools.fs import RecordFs
+from aws_doc_sdk_examples_tools.lliam.domain.commands import CreatePrompts
+from aws_doc_sdk_examples_tools.lliam.service_layer.create_prompts import create_prompts
+from aws_doc_sdk_examples_tools.lliam.service_layer.unit_of_work import FsUnitOfWork
+
+
+def test_create_prompts_writes_when_commit_called():
+    """Test that create_prompts successfully writes prompts when commit is called."""
+    fs = RecordFs(
+        {
+            Path("/system1.md"): "System prompt 1 content",
+            Path("/system2.md"): "System prompt 2 content",
+            Path("/fake/doc_gen_root"): "",
+        }
+    )
+    uow = FsUnitOfWork(fs=fs)
+    cmd = CreatePrompts(
+        doc_gen_root="/fake/doc_gen_root",
+        system_prompts=["/system1.md", "/system2.md"],
+        out_dir="/fake/output",
+    )
+
+    create_prompts(cmd, uow)
+
+    # Ailly config should be in committed prompts
+    ailly_config_path = Path("/fake/output/.aillyrc")
+    assert fs.stat(ailly_config_path).exists
+    ailly_config_content = fs.read(ailly_config_path)
+    assert "System prompt 1 content" in ailly_config_content
+    assert "System prompt 2 content" in ailly_config_content