V0.1.4

pyproject.toml

@@ -7,15 +7,15 @@ packages=["ragdaemon"]
 
 [project]
 name = "ragdaemon"
-version = "0.1.3"
+version = "0.1.4"
 description = "Generate and render a call graph for a Python project."
 readme = "README.md"
 dependencies = [
     "chromadb==0.4.24",
     "fastapi==0.109.2",
     "Jinja2==3.1.3",
     "networkx==3.2.1",
-    "spiceai==0.1.9",
+    "spiceai==0.1.11",
     "starlette==0.36.3",
     "tiktoken==0.6.0",
     "tqdm==4.66.2",

ragdaemon/__init__.py

@@ -1 +1 @@
-__version__ = "0.1.3"
+__version__ = "0.1.4"

ragdaemon/annotators/__init__.py

@@ -1,7 +1,7 @@
 from ragdaemon.annotators.base_annotator import Annotator  # noqa: F401
 from ragdaemon.annotators.chunker import Chunker
-from ragdaemon.annotators.chunker_llm import ChunkerLLM
 from ragdaemon.annotators.chunker_line import ChunkerLine
+from ragdaemon.annotators.chunker_llm import ChunkerLLM
 from ragdaemon.annotators.diff import Diff
 from ragdaemon.annotators.hierarchy import Hierarchy
 from ragdaemon.annotators.layout_hierarchy import LayoutHierarchy

ragdaemon/annotators/diff.py

@@ -3,15 +3,15 @@
 from pathlib import Path
 
 import networkx as nx
+from spice import Spice
 
 from ragdaemon.annotators.base_annotator import Annotator
 from ragdaemon.database import (
     DEFAULT_EMBEDDING_MODEL,
-    Database,
     MAX_TOKENS_PER_EMBEDDING,
+    Database,
 )
 from ragdaemon.errors import RagdaemonError
-from ragdaemon.llm import token_counter
 from ragdaemon.utils import get_document, hash_str, parse_path_ref
 
 
@@ -109,9 +109,7 @@ async def annotate(
 
             # If the full diff is too long to embed, it is truncated. Anything
             # removed will be captured in chunks.
-            tokens = token_counter(
-                document, model=DEFAULT_EMBEDDING_MODEL, full_message=False
-            )
+            tokens = Spice().count_tokens(document, model=DEFAULT_EMBEDDING_MODEL)
             if tokens > MAX_TOKENS_PER_EMBEDDING:
                 truncate_ratio = (MAX_TOKENS_PER_EMBEDDING / tokens) * 0.99
                 document = document[: int(len(document) * truncate_ratio)]

ragdaemon/annotators/hierarchy.py

@@ -2,11 +2,12 @@
 from pathlib import Path
 
 import networkx as nx
+from spice import Spice
 
 from ragdaemon.annotators.base_annotator import Annotator
 from ragdaemon.database import MAX_TOKENS_PER_EMBEDDING, Database
 from ragdaemon.errors import RagdaemonError
-from ragdaemon.llm import token_counter
+from ragdaemon.llm import DEFAULT_COMPLETION_MODEL
 from ragdaemon.utils import get_document, get_non_gitignored_files, hash_str
 
 
@@ -56,7 +57,7 @@ def get_active_checksums(
             path_str = path.as_posix()
             ref = path_str
             document = get_document(ref, cwd)
-            tokens = token_counter(document)
+            tokens = Spice().count_tokens(document, DEFAULT_COMPLETION_MODEL)
             if tokens > MAX_TOKENS_PER_EMBEDDING:  # e.g. package-lock.json
                 continue
             checksum = hash_str(document)

ragdaemon/app.py

@@ -12,9 +12,9 @@
 from spice import Spice
 from starlette.templating import Jinja2Templates
 
-from ragdaemon.llm import DEFAULT_COMPLETION_MODEL
-from ragdaemon.database import DEFAULT_EMBEDDING_MODEL
 from ragdaemon.daemon import Daemon
+from ragdaemon.database import DEFAULT_EMBEDDING_MODEL
+from ragdaemon.llm import DEFAULT_COMPLETION_MODEL
 
 # Load daemon with command line arguments and visualization annotators
 parser = argparse.ArgumentParser(description="Start the ragdaemon server.")

ragdaemon/daemon.py

@@ -9,8 +9,8 @@
 
 from ragdaemon.annotators import Annotator, annotators_map
 from ragdaemon.context import ContextBuilder
-from ragdaemon.database import Database, DEFAULT_EMBEDDING_MODEL, get_db
-from ragdaemon.llm import DEFAULT_COMPLETION_MODEL, token_counter
+from ragdaemon.database import DEFAULT_EMBEDDING_MODEL, Database, get_db
+from ragdaemon.llm import DEFAULT_COMPLETION_MODEL
 from ragdaemon.utils import get_non_gitignored_files
 
 
@@ -128,7 +128,9 @@ def get_context(
             # TODO: Compare graph hashes, reconcile changes
             context = context_builder
         include_context_message = context.render()
-        include_tokens = token_counter(include_context_message)
+        include_tokens = self.spice_client.count_tokens(
+            include_context_message, DEFAULT_COMPLETION_MODEL
+        )
         if not auto_tokens or include_tokens >= max_tokens:
             return context
 
@@ -140,7 +142,9 @@ def get_context(
             else:
                 context.add_ref(node["ref"], tags=["search-result"])
             next_context_message = context.render()
-            next_tokens = token_counter(next_context_message)
+            next_tokens = self.spice_client.count_tokens(
+                next_context_message, DEFAULT_COMPLETION_MODEL
+            )
             if (next_tokens - include_tokens) > auto_tokens:
                 if node["type"] == "diff":
                     context.remove_diff(node["id"])

ragdaemon/database/__init__.py

@@ -1,10 +1,11 @@
 import os
 from pathlib import Path
 
-from spice import Spice, SpiceError
+from spice import Spice
+from spice.errors import SpiceError
 
-from ragdaemon.database.database import Database
 from ragdaemon.database.chroma_database import ChromaDB
+from ragdaemon.database.database import Database
 from ragdaemon.database.lite_database import LiteDB
 from ragdaemon.utils import mentat_dir_path
 

ragdaemon/llm.py

@@ -1,12 +1 @@
-import tiktoken
-
-
 DEFAULT_COMPLETION_MODEL = "gpt-4-0125-preview"
-
-
-def token_counter(
-    text: str, model: str | None = None, full_message: bool = False
-) -> int:
-    if model is None:
-        model = DEFAULT_COMPLETION_MODEL
-    return len(tiktoken.encoding_for_model(model).encode(text))

ragdaemon/utils.py

@@ -6,7 +6,6 @@
 
 from ragdaemon.errors import RagdaemonError
 
-
 mentat_dir_path = Path.home() / ".mentat"
 
 

requirements.txt

@@ -5,7 +5,7 @@ Jinja2==3.1.3
 networkx==3.2.1
 pytest==8.0.2
 pytest-asyncio==0.23.5
-spiceai==0.1.9
+spiceai==0.1.11
 starlette==0.36.3
 tiktoken==0.6.0
 tqdm==4.66.2

tests/conftest.py

@@ -1,3 +1,4 @@
+import os
 import shutil
 import subprocess
 import tempfile
@@ -64,3 +65,9 @@ def git_history(cwd):
             f.write("print('Hello, world!')\n")
 
         yield tmpdir_path
+
+
+# We have to set the key since counting tokens with an openai model loads the openai client
+@pytest.fixture(autouse=True)
+def mock_openai_api_key():
+    os.environ["OPENAI_API_KEY"] = "fake_key"