Hermes / Local LLMs

You have access to Alchemist — a financial research database of primary source documents
(FOMC minutes, USDA WASDE, SEC filings, BLS/BEA data releases, and more).
Every result includes verbatim source quotes and publication dates.

## Available Tools

search_insights(query, source_name?, since?, limit?)
  Primary search. Use for any financial research question.
  Returns citable insights with verbatim excerpts.

get_document(doc_id)
  Fetch a full document and all its extracted insights.
  Use doc_ids returned from search_insights.

list_documents(source_name?, since?, limit?)
  Browse available documents by source or date.

list_filter_options()
  Return all valid source_name values.
  Call before filtering by source.

## API

Base URL: https://api.askalchemist.com
Auth header: X-API-Key: YOUR_KEY

search_insights endpoint:
  POST /api/insights/search
  Body: {"query": "...", "source_name": null, "since": null, "limit": 20}

get_document endpoint:
  GET /api/documents/{doc_id}

list_documents endpoint:
  GET /api/documents?source_name=...&since=...&limit=50

list_filter_options endpoint:
  GET /api/documents/filters

## Query Format (Critical)

Phrase queries as ONE focused concept — a concise question or 3–8 keyword phrase.

Good:
- "Fed dot plot rate path 2026"
- "AAPL Q3 2025 revenue guidance"
- "corn ending stocks USDA WASDE"

Bad:
- "Tell me about the Fed" (too vague)
- "Fed rates inflation corn energy bonds" (keyword soup)

## Workflow

1. Call list_filter_options if you need to filter by source
2. Run 1–2 targeted searches, read results, then refine
3. If sparse results: retry with different phrasing, remove filters
4. For deep research: search for doc_id → call get_document

## Citation Format

Always cite source_name and published_date:
"Per the Federal Reserve FOMC Minutes (March 2026)..."

import ollama
import httpx
import json

ALCHEMIST_KEY = "YOUR_KEY"

tools = [
    {
        "type": "function",
        "function": {
            "name": "search_insights",
            "description": "Search the Alchemist financial research database. Returns citable insights with verbatim source excerpts. Query as a focused 3–8 word phrase or concise question.",
            "parameters": {
                "type": "object",
                "properties": {
                    "query": {"type": "string"},
                    "source_name": {"type": "string"},
                    "since": {"type": "string", "description": "ISO date YYYY-MM-DD"},
                    "limit": {"type": "integer", "default": 20},
                },
                "required": ["query"],
            },
        },
    },
    {
        "type": "function",
        "function": {
            "name": "list_filter_options",
            "description": "Return all valid source_name values. Call before filtering searches.",
            "parameters": {"type": "object", "properties": {}, "required": []},
        },
    },
    {
        "type": "function",
        "function": {
            "name": "get_document",
            "description": "Fetch a full document and all its insights by doc_id.",
            "parameters": {
                "type": "object",
                "properties": {"doc_id": {"type": "string"}},
                "required": ["doc_id"],
            },
        },
    },
]


def call_alchemist(name, args):
    h = {"X-API-Key": ALCHEMIST_KEY}
    if name == "search_insights":
        return httpx.post("https://api.askalchemist.com/api/insights/search", headers=h, json=args).json()
    if name == "list_filter_options":
        return httpx.get("https://api.askalchemist.com/api/documents/filters", headers=h).json()
    if name == "get_document":
        return httpx.get(f"https://api.askalchemist.com/api/documents/{args['doc_id']}", headers=h).json()
    return {"error": f"unknown tool {name}"}


messages = [
    {"role": "system", "content": "You have access to Alchemist — a financial research database. Use search_insights as your primary tool. Query as a focused 3–8 word phrase. Always cite source_name and published_date."},
    {"role": "user", "content": "What did the March 2026 FOMC minutes say about inflation?"},
]

while True:
    response = ollama.chat(model="hermes3:70b", messages=messages, tools=tools)
    msg = response.message
    messages.append(msg)

    if not msg.tool_calls:
        print(msg.content)
        break

    for tc in msg.tool_calls:
        result = call_alchemist(tc.function.name, tc.function.arguments)
        messages.append({"role": "tool", "content": json.dumps(result)})

from openai import OpenAI

client = OpenAI(base_url="http://localhost:1234/v1", api_key="lm-studio")
# Use the same tools list and call_alchemist function above
response = client.chat.completions.create(
    model="NousResearch/Hermes-3-Llama-3.1-70B",
    tools=tools,
    messages=messages,
)