Vertex update to Main

ABB-Manual-Assistant/README.md

@@ -0,0 +1,3 @@
+# Linamar-Vector-Bootcamp
+
+## Applying agentic ai to robot troubleshooting.

ABB-Manual-Assistant/Run_Eval_Info.txt

@@ -0,0 +1,5 @@
+Upload dataset command:
+python3 upload_test_data_from_file.py --dataset-name "SME Test Set" --input-file SME-TestSet.csv
+
+Run evaluation command:
+python3 run_eval_from_dataset.py --dataset-name "SME Test Set" --run-name "SME Eval Run"

ABB-Manual-Assistant/agent_utils/memory_store.py

@@ -0,0 +1,67 @@
+import datetime
+import os
+import sqlite3
+import uuid
+
+
+class MemoryStore:
+    def __init__(self, db_path="data/memory.db"):
+        os.makedirs(os.path.dirname(db_path), exist_ok=True)
+        self.conn = sqlite3.connect(db_path, check_same_thread=False)
+        self._init_db()
+
+    def _init_db(self):
+        cur = self.conn.cursor()
+        cur.execute("""
+        CREATE TABLE IF NOT EXISTS conversations (
+            id TEXT PRIMARY KEY,
+            title TEXT,
+            created_at TEXT
+        )
+        """)
+        cur.execute("""
+        CREATE TABLE IF NOT EXISTS messages (
+            id INTEGER PRIMARY KEY AUTOINCREMENT,
+            conversation_id TEXT,
+            role TEXT,
+            content TEXT,
+            timestamp TEXT
+        )
+        """)
+        self.conn.commit()
+
+    def create_conversation(self, title=None):
+        cid = str(uuid.uuid4())
+        if not title:
+            title = f"Chat {datetime.datetime.now().strftime('%Y-%m-%d %H:%M')}"
+        self.conn.execute(
+            "INSERT INTO conversations VALUES (?, ?, ?)", (cid, title, datetime.datetime.now().isoformat())
+        )
+        self.conn.commit()
+        return cid
+
+    def list_conversations(self):
+        return self.conn.execute("SELECT id, title, created_at FROM conversations ORDER BY created_at DESC").fetchall()
+
+    def get_history(self, conversation_id, limit=100):
+        rows = self.conn.execute(
+            "SELECT role, content FROM messages WHERE conversation_id=? ORDER BY id ASC LIMIT ?",
+            (conversation_id, limit),
+        ).fetchall()
+        return [{"role": r[0], "content": r[1]} for r in rows]
+
+    def log_message(self, conversation_id, role, content):
+        self.conn.execute(
+            "INSERT INTO messages (conversation_id, role, content, timestamp) VALUES (?, ?, ?, ?)",
+            (conversation_id, role, content, datetime.datetime.now().isoformat()),
+        )
+        self.conn.commit()
+
+    def rename_conversation(self, conversation_id, new_title):
+        self.conn.execute("UPDATE conversations SET title=? WHERE id=?", (new_title, conversation_id))
+        self.conn.commit()
+
+    def delete_conversation(self, conversation_id):
+        self.conn.execute("DELETE FROM messages WHERE conversation_id=?", (conversation_id,))
+        self.conn.execute("DELETE FROM conversations WHERE id=?", (conversation_id,))
+        self.conn.commit()

ABB-Manual-Assistant/configs.py

@@ -0,0 +1,211 @@
+"""Configuration settings for agent evaluations.
+
+This module provides centralized configuration management using Pydantic settings,
+supporting environment variables and .env file loading.
+"""
+
+from typing import Any
+
+from pydantic import AliasChoices, BaseModel, Field, SecretStr, field_validator
+from pydantic_settings import BaseSettings, SettingsConfigDict
+from sqlalchemy.engine.url import URL
+
+
+class DatabaseConfig(BaseModel):
+    """Database connection configuration."""
+
+    driver: str = Field(
+        ...,
+        description="SQLAlchemy dialect (e.g., 'sqlite', 'postgresql', 'mysql+pymysql').",
+    )
+    username: str | None = Field(
+        default=None,
+        description="Database username. For SQLite or integrated authentication, this can be None.",
+    )
+    host: str | None = Field(default=None, description="Database host address or file path for SQLite.")
+    password: SecretStr | None = Field(
+        default=None,
+        description="Database password. For SQLite or integrated authentication, this can be None.",
+    )
+    port: int | None = Field(default=None, description="Database port number.")
+    database: str | None = Field(default=None, description="Database name or file path for SQLite.")
+    query: dict[str, Any] = Field(
+        default_factory=dict,
+        description="URL query parameters (e.g. {'mode': 'ro'} for read-only SQLite).",
+    )
+
+    def build_uri(self) -> str:
+        """Construct the SQLAlchemy connection URI safely using the official URL object.
+
+        This handles special character escaping in passwords automatically.
+
+        Returns
+        -------
+        str
+            The full database connection URI.
+        """
+        return URL.create(
+            drivername=self.driver,
+            username=self.username,
+            password=self.password.get_secret_value() if self.password else None,
+            host=self.host,
+            port=self.port,
+            database=self.database,
+            query=self.query,
+        ).render_as_string(hide_password=False)
+
+
+class Configs(BaseSettings):
+    """Central configuration for all agent evaluations.
+
+    This class automatically loads configuration values from environment variables
+    and a .env file. Service-specific fields are optional - agents validate
+    required fields at initialization.
+
+    Examples
+    --------
+    >>> from aieng.agent_evals.configs import Configs
+    >>> config = Configs()
+    >>> print(config.default_worker_model)
+    'gemini-2.5-flash'
+    """
+
+    model_config = SettingsConfigDict(
+        env_file=".env",
+        env_file_encoding="utf-8",
+        env_ignore_empty=True,
+        env_nested_delimiter="__",
+        extra="ignore",
+    )
+
+    aml_db: DatabaseConfig | None = Field(
+        default=None,
+        description="Anti-Money Laundering database configuration. Used by the Fraud Investigation Agent.",
+    )
+
+    report_generation_db: DatabaseConfig | None = Field(
+        default=None,
+        description="Database configuration for the the Report Generation Agent.",
+    )
+
+    # === Core LLM Settings ===
+    openai_base_url: str = Field(
+        default="https://generativelanguage.googleapis.com/v1beta/openai/",
+        description="Base URL for OpenAI-compatible API (defaults to Gemini endpoint).",
+    )
+    openai_api_key: SecretStr = Field(
+        validation_alias=AliasChoices("OPENAI_API_KEY", "GEMINI_API_KEY", "GOOGLE_API_KEY"),
+        description="API key for OpenAI-compatible API (accepts OPENAI_API_KEY, GEMINI_API_KEY, or GOOGLE_API_KEY).",
+    )
+    google_api_key: SecretStr = Field(
+        validation_alias=AliasChoices("GEMINI_API_KEY", "GOOGLE_API_KEY"),
+        description="API key for Google/Gemini API (accepts GEMINI_API_KEY or GOOGLE_API_KEY).",
+    )
+    anthropic_api_key: SecretStr | None = Field(
+        default=None,
+        validation_alias="ANTHROPIC_API_KEY",
+        description="API key for Anthropic API access when using LiteLLM-backed Claude models.",
+    )
+    vector_inference_api_key: SecretStr | None = Field(
+        default=None,
+        validation_alias="VECTOR_INFERENCE_API_KEY",
+        description="API key for Vector's internal OpenAI-compatible inference endpoint.",
+    )
+    default_planner_model: str = Field(
+        default="gemini-2.5-pro",
+        description="Model name for planning/complex reasoning tasks.",
+    )
+    default_worker_model: str = Field(
+        default="gemini-2.5-flash",
+        description="Model name for worker/simple tasks.",
+    )
+    default_evaluator_model: str = Field(
+        default="gemini-2.5-pro",
+        description="Model name for LLM-as-judge evaluation tasks.",
+    )
+    default_temperature: float = Field(
+        default=1.0,
+        ge=0.0,
+        le=2.0,
+        description="Default temperature for LLM generation. Lower values (0.0-0.3) produce more consistent outputs.",
+    )
+    default_evaluator_temperature: float = Field(
+        default=0.0,
+        ge=0.0,
+        le=2.0,
+        description="Temperature for LLM-as-judge evaluations. Default 0.0 for deterministic judging.",
+    )
+
+    # === Tracing (Langfuse) ===
+    langfuse_public_key: str | None = Field(
+        default=None,
+        pattern=r"^pk-lf-.*$",
+        description="Langfuse public key for tracing (must start with 'pk-lf-').",
+    )
+    langfuse_secret_key: SecretStr | None = Field(
+        default=None,
+        description="Langfuse secret key for tracing (must start with 'sk-lf-').",
+    )
+    langfuse_host: str = Field(
+        default="https://us.cloud.langfuse.com",
+        validation_alias="LANGFUSE_HOST",
+        description="Langfuse base URL.",
+    )
+
+    # === Embedding Service ===
+    embedding_base_url: str | None = Field(default=None, description="Base URL for embedding API service.")
+    embedding_api_key: SecretStr | None = Field(default=None, description="API key for embedding service.")
+    embedding_model_name: str = Field(default="@cf/baai/bge-m3", description="Name of the embedding model.")
+
+    # === E2B Code Interpreter ===
+    e2b_api_key: SecretStr | None = Field(
+        default=None,
+        description="E2B.dev API key for code interpreter (must start with 'e2b_').",
+    )
+    default_code_interpreter_template: str | None = Field(
+        default="9p6favrrqijhasgkq1tv",
+        description="Default template name or ID for E2B.dev code interpreter.",
+    )
+
+    # === Web Search ===
+    web_search_base_url: str | None = Field(default=None, description="Base URL for web search service.")
+    web_search_api_key: SecretStr | None = Field(default=None, description="API key for web search service.")
+
+    # === Vertex AI Search (custom knowledge base) ===
+    google_cloud_location: str = Field(
+        default="us-central1",
+        description="GCP region for Vertex AI model calls. Must match a region that supports Gemini.",
+    )
+    vertex_datastore_id: str | None = Field(
+        default=None,
+        validation_alias="VERTEX_AI_DATASTORE_ID",
+        description=(
+            "Full Vertex AI Search data store resource name. "
+            "Format: projects/{project}/locations/global/collections/default_collection/dataStores/{id}. "
+            "Authentication uses Application Default Credentials (ADC) — no API key required."
+        ),
+    )
+
+    # === Report Generation ===
+    # Defaults are set in the implementations/report_generation/env_vars.py file
+    report_generation_output_path: str | None = Field(
+        default=None,
+        description="Path to the directory where the report generation agent will save the reports.",
+    )
+
+    # Validators for the SecretStr fields
+    @field_validator("langfuse_secret_key")
+    @classmethod
+    def validate_langfuse_secret(cls, v: SecretStr | None) -> SecretStr | None:
+        """Validate that the Langfuse secret key starts with 'sk-lf-'."""
+        if v is not None and not v.get_secret_value().startswith("sk-lf-"):
+            raise ValueError("Langfuse secret key must start with 'sk-lf-'")
+        return v
+
+    @field_validator("e2b_api_key")
+    @classmethod
+    def validate_e2b_key(cls, v: SecretStr | None) -> SecretStr | None:
+        """Validate that the E2B API key starts with 'e2b_' if provided."""
+        if v is not None and not v.get_secret_value().startswith("e2b_"):
+            raise ValueError("E2B API key must start with 'e2b_'")
+        return v

ABB-Manual-Assistant/conversation_manager.py

@@ -0,0 +1,73 @@
+# agents/conversation_manager.py
+from __future__ import annotations
+
+from typing import Any, Dict, List, Optional
+
+from agent_utils.memory_store import MemoryStore
+
+
+class ConversationManagerAgent:
+    """
+    Minimal 'agent' wrapper around MemoryStore so other agents/tools
+    can call a stable interface. Keeps things simple:
+    - make sure a conversation exists
+    - append user/assistant turns
+    - read history
+    - rename/delete/list conversations
+    - optional: keep assistant partials in RAM; persist only on finalize
+    """
+
+    def __init__(self, store: Optional[MemoryStore] = None):
+        self.store = store or MemoryStore()
+        # simple in-RAM buffer for current assistant partials per conversation
+        self._partials: Dict[str, str] = {}
+
+    # ---- conversation mgmt ----
+    def ensure_conversation(self, conversation_id: Optional[str]) -> str:
+        rows = self.store.list_conversations()
+        if not rows:
+            return self.store.create_conversation()
+        if conversation_id:
+            return conversation_id
+        # default to newest (list_conversations should return DESC by created_at)
+        return rows[0][0]
+
+    def list_conversations(self) -> List[Dict[str, Any]]:
+        rows = self.store.list_conversations()
+        return [{"id": r[0], "title": r[1], "created_at": r[2]} for r in rows]
+
+    def rename(self, conversation_id: str, title: str) -> Dict[str, Any]:
+        self.store.rename_conversation(conversation_id, title)
+        return {"ok": True, "conversation_id": conversation_id, "title": title}
+
+    def delete(self, conversation_id: str) -> Dict[str, Any]:
+        self.store.delete_conversation(conversation_id)
+        self._partials.pop(conversation_id, None)
+        return {"ok": True}
+
+    def create(self, title: Optional[str] = None) -> Dict[str, Any]:
+        cid = self.store.create_conversation(title)
+        return {"ok": True, "conversation_id": cid}
+
+    # ---- messages ----
+    def save_user(self, conversation_id: str, content: str) -> Dict[str, Any]:
+        cid = self.ensure_conversation(conversation_id)
+        self.store.log_message(cid, "user", content)
+        return {"ok": True, "conversation_id": cid}
+
+    def set_assistant_partial(self, conversation_id: str, partial_text: str) -> Dict[str, Any]:
+        # Keep partials in RAM for simplicity/quickness
+        cid = self.ensure_conversation(conversation_id)
+        self._partials[cid] = partial_text
+        return {"ok": True, "conversation_id": cid}
+
+    def finalize_assistant(self, conversation_id: str) -> Dict[str, Any]:
+        cid = self.ensure_conversation(conversation_id)
+        final_text = self._partials.pop(cid, "")
+        # Only persist once at the end of the stream
+        self.store.log_message(cid, "assistant", final_text)
+        return {"ok": True, "conversation_id": cid, "content_len": len(final_text)}
+
+    def get_history_messages(self, conversation_id: str, limit: int = 1000) -> List[Dict[str, Any]]:
+        cid = self.ensure_conversation(conversation_id)
+        return self.store.get_history(cid, limit=limit)