This is a submission for the Hermes Agent Challenge: Build With Hermes Agent
The Problem Nobody Talks About
Every time the Reserve Bank of India publishes a circular, somewhere inside an Indian bank, a compliance officer opens a PDF.
They read it. They try to figure out what it means for their institution specifically. They write a summary email. They forward it to five department heads. They chase those department heads for two weeks to confirm it's been actioned. They build a spreadsheet to track all of this. And then the next circular drops and the cycle starts again.
RBI publishes hundreds of circulars a year. SEBI publishes more. MCA publishes more still. Compliance teams at Indian banks are drowning — not because they're incompetent, but because the volume of regulatory output has outpaced any reasonable human ability to track it manually.
The fine for missing a deadline isn't a polite reminder. It's a penalty notice.
This is the problem I built for.
What I Built
RBI Regulatory Digest Agent — an autonomous multi-step agent powered by Hermes Agent that monitors RBI and SEBI publication feeds, reads every new circular, extracts structured action points from the regulatory text, and delivers a formatted intelligence report to compliance teams automatically.
No human reads the circular first. No human decides what's important. No human routes it to the right department. The agent does all of that.
The pipeline
RBI/SEBI feeds → new circular detected → full text extracted →
LLM analysis → structured action points → risk classification →
HTML dashboard generated → email delivered
Every action point extracted contains:
- What needs to be done — specific and actionable, not a vague summary
- Deadline — parsed from the circular text
- Responsible department — Credit, Compliance, Treasury, Operations, IT, Legal
- Evidence required — what documentation confirms completion
- Priority — Critical (overdue or <7 days), High, Medium, Low
From a new circular to a structured compliance briefing in under 10 seconds.
Demo
The demo shows Hermes Agent receiving a single instruction and autonomously:
- Polling the RBI notification feed
- Detecting 10 new circulars
- Fetching and parsing each circular's text
- Running LLM extraction on each one
- Generating a risk-classified HTML dashboard with 46 action points across 8 departments
- Delivering the digest by email
No human intervention between step 1 and step 6.
Code
GitHub: https://github.com/kunalp-singh/rbi-digest
Project structure
rbi_monitor.py # Polls RBI/SEBI RSS feeds, deduplicates, stores to SQLite
extractor.py # Fetches circular text, runs LLM extraction, stores action points
digest.py # Builds risk-classified HTML dashboard, sends email
hermes_prompt.txt # The single instruction Hermes receives to run everything
circulars.db # SQLite store (gitignored)
My Tech Stack
- Hermes Agent — orchestrates the full pipeline via terminal tool
- Python 3.11 — core runtime
- Ollama + Qwen3:8b — local LLM for circular analysis, runs entirely on-device
- BeautifulSoup4 + lxml — RSS feed parsing
- SQLite — circular deduplication and action point storage
- OpenAI-compatible client — connects Python to local Ollama endpoint
- SMTP / Gmail — email delivery
- HTML/CSS — dashboard with risk heatmap, compliance timeline, department cards
Everything runs locally. No data leaves the machine. No cloud LLM API required.
How I Used Hermes Agent
This is the part that matters most for the challenge, so I want to be precise about it.
The single instruction
The entire pipeline runs from one message to Hermes:
Use the terminal tool to run this command exactly:
cd ~/Studies/Projects/hermesProject && source venv/bin/activate &&
python3 rbi_monitor.py && python3 extractor.py && python3 digest.py
Hermes takes this, uses its terminal tool to execute the full chain, monitors the output of each step, and reports back a summary — how many circulars were found, how many action points extracted, which departments have critical items, and confirmation that the digest was delivered.
Why Hermes was the right fit
I could have run a cron job directly. I chose Hermes for three specific reasons:
1. Multi-step reasoning across tool calls
Each script depends on the previous one succeeding. Hermes doesn't just fire and forget — it monitors output, detects failures, and can adapt. When the extractor found 0 action points on informational circulars, Hermes correctly reported that as expected behaviour rather than an error.
2. Natural language orchestration
The hermes_prompt.txt file is plain English. Any compliance team member can read it, understand what the agent is doing, and modify it. There's no bash scripting knowledge required to change which feeds are monitored or what gets reported.
3. Episodic memory for future runs
Hermes logs what it processed in each session. Future runs can reference past extractions — "you processed this circular last week, here's what changed" — which is exactly the kind of institutional memory a compliance team needs.
Agentic capabilities used
- Terminal tool — executes the monitoring, extraction, and digest pipeline
- Multi-step execution — coordinates three dependent scripts in sequence
- Output interpretation — reads script output and generates a human-readable summary
- Planning — determines correct execution order without being told explicitly
What the Output Looks Like
From the most recent run against live RBI feeds:
- 46 action points extracted from 20 circulars
- 8 departments impacted: Compliance, Credit, Treasury, IT, Operations, Legal, Technology, Board/Governance
- 11 critical items — deadlines already passed or within 7 days
- 20 items with specific upcoming deadlines tracked in the compliance timeline
Sample action points extracted by the LLM:
Compliance dept — CRITICAL: Submit required documents for claiming Agency Commission — Deadline: April 30, 2026 — Source: RBI Conduct of Government Business Directions
Credit dept — MEDIUM: Update credit risk assessment policies to include calamity impact on borrowers — Deadline: July 1, 2026 — Source: RBI Credit Risk Management Amendment Directions
Operations dept — MEDIUM: Appoint Nodal Officers for pension grievance handling — Deadline: Immediate effect — Source: RBI Disbursement of Government Pension Directions
Every item links directly to the source RBI circular.
What's Next
This is a working prototype, not a finished product. The honest limitations:
- Currently monitors RBI and SEBI only — IRDAI, MCA, FATF feeds can be added
- LLM extraction runs locally on Qwen3:8b — a larger model would improve accuracy on dense legal text
- The dashboard is static HTML — a real deployment would have a live web interface with team login, action acknowledgement, and audit trail
The architecture is modular enough that each of these is an extension, not a rewrite.
The core insight this project validates: regulatory compliance monitoring is exactly the kind of repetitive, high-stakes, structured reasoning task that agentic AI should be doing instead of humans. Hermes Agent made it possible to prototype that in days rather than months.
Built during the Hermes Agent Challenge, May 2026. All data sourced from live RBI publication feeds.
Top comments (0)