Dhaara

Dhaara (धारा) — a stream, a current, a flow.

Your personal journal, as a conversation. Talk to it on Telegram — by voice or text, in English or any Indian language — and Dhaara quietly turns your thoughts into a well-organised, searchable archive of your life. No apps to open. No forms to fill. Just talk.

Dhaara is the first agent in the PAI (Personal AI) ecosystem — a set of agents that share a common data store and, over time, learn to help you across work, health, finances, and growth.

Why Dhaara?

Most journaling apps ask you to sit down, open a page, and write. Most of us don't. Life happens in the gaps — in the auto ride, between meetings, while making chai. Dhaara lives in the one app you already check a hundred times a day: Telegram.

🎙️ Voice-first. Ramble in Hindi, Tamil, Marathi, English — or a mix. Dhaara understands.
🧠 Agentic, not a form. It classifies each entry, extracts mood and a free-form subcategory tag, and can edit or delete past entries when you ask.
📒 One file per day. Every day is a single markdown file with four sections — WORK, PERSONAL, HABITS, FINANCE — and timestamped bullet entries inside each.
🏷️ Inline tags for free. Every bullet carries its own [time] [CATEGORY/subcategory] prefix and optional *(mood: ...)*, so each line is a self-contained chunk that's perfect for RAG later.
💰 Automatic expense capture. "Spent 30 on milk, 150 on auto" lands as entries under FINANCE/groceries and FINANCE/transport.
🎯 TELOS-aware. You give it your goals and priorities in _telos/ and the agent reads entries in that context.
🗂️ Your data, your disk. Plain markdown files. No cloud lock-in. No database. grep just works.
🔓 BYO-model. Runs on AWS Bedrock or OpenRouter — pick Claude, Nova, Mistral, Gemini, Llama, whatever you like.
👤 Single-user by design. The bot answers to exactly one Telegram ID — yours.
📊 Analytics out of the box. Streak nudges, activity heatmaps, mood timelines, weekly summaries, search with context, and a combined dashboard — all reading the same plain markdown. See § Tools.

How it feels

you:   spent 30 on milk, 20 on cold drink, 150 on auto today
bot:   Recorded to [FINANCE].
       (three bullets added under FINANCE/groceries and FINANCE/transport)

you:   [voice note in Hindi] aaj kaafi thaka hua hoon, kaam pe focus nahi ho paa raha
bot:   Recorded to [PERSONAL]. (mood: tired)

you:   worked on dhaara for 2 hrs, phase 1 is nearly done
bot:   Recorded to [WORK].

you:   what did I eat today?
bot:   L12: - [1:30 PM] [FINANCE/food] Spent ₹150 on lunch
       L18: - [8:10 PM] [FINANCE/food] Ordered dinner from Swiggy ₹420

you:   delete the swiggy one
bot:   Deleted: - [8:10 PM] [FINANCE/food] Ordered dinner from Swiggy ₹420

Every entry is appended to today's single markdown file under the right section. Nothing hidden. Nothing proprietary. Every bullet is timestamped and tagged inline, so the file itself is both a journal and an index.

The agent can also reflect over time when you ask it to — see § Asking for TELOS insights for how it grounds those answers in your actual journal data.

Quick start

You need:

Python 3.11+
A Telegram bot token — 2 minutes via @BotFather
Your Telegram user ID — from @userinfobot
A Sarvam AI API key — sarvam.ai (for voice + Indian-language translation; free tier works)
One of:
- An OpenRouter API key — openrouter.ai/keys (easiest path — credit card, one model ID, done)
- An AWS account with Bedrock access (if you prefer to run on your own cloud)

1. Clone and install

git clone https://github.com/noxtandav/Dhaara.git
cd Dhaara

python3 -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install -r requirements.txt

2. Configure

cp config.example.yaml config.yaml

Open config.yaml and fill in your values. The file is gitignored — it never leaves your machine.

Minimal OpenRouter setup:

telegram:
  bot_token: "123456:ABC..."
  authorized_user_id: 123456789

data_dir: "~/PAI/DhaaraData"
timezone: "Asia/Kolkata"

ai:
  provider: "openrouter"

openrouter:
  model: "anthropic/claude-sonnet-4.5"
  api_key: "sk-or-v1-..."

sarvam:
  api_key: "..."

Or use Bedrock:

ai:
  provider: "bedrock"

bedrock:
  model_id: "us.amazon.nova-pro-v1:0"
  region: "us-east-1"
  aws_profile: "default"

3. Verify your config (optional but recommended)

python scripts/check_config.py

A static linter that catches the things that bite first-time users — leftover placeholders from config.example.yaml, missing keys for the chosen AI provider, invalid timezones, oddly-shaped credentials. Exits non-zero on errors so you can wire it into a pre-deploy step. Add -f json for machine-readable output.

4. Initialize the data layout

python scripts/init.py            # creates data_dir/, journal/, _telos/ + seeds example TELOS files
python scripts/init.py --dry-run  # see what would happen first

Idempotent — re-running tells you what was already there and never overwrites your TELOS files. Pairs naturally with check_config.py: the linter validates your config, init.py realizes the file system that config points at.

5. Run

# Foreground (for testing)
python -m src.main

Message your bot — that's it. For production deployment with PM2 and the auto-scheduled review jobs, see § Running with PM2 below.

Running with PM2

For long-running deployments, dhaara ships with a PM2 manifest that registers four apps in one go: the always-on bot plus three scheduled review jobs.

App	What it does	Schedule
`dhaara`	The Telegram bot	Always on, auto-restarts on failure
`dhaara-weekly`	Markdown weekly review → `<data_dir>/weekly/YYYY-Www.md`	Sundays 21:07 local
`dhaara-dashboard`	Rolling 7-day combined dashboard → `<data_dir>/dashboards/YYYY-MM-DD.md`	Daily 23:53 local
`dhaara-streak`	Desktop notification if your streak broke (macOS `osascript`)	Daily 19:37 local

Schedules use off-the-round minutes (:07, :53, :37) to dodge the fleet-wide thundering herd of 0 21 / 0 0 / 0 19.

One-time setup

npm install -g pm2
cp ecosystem.config.js.example ecosystem.config.js

Open ecosystem.config.js and edit the two paths near the top:

dataDir — should match your config.yaml's data_dir
logsDir — where PM2 writes process logs (default ./logs, gitignored)

Start and persist

pm2 start ecosystem.config.js   # registers all four apps
pm2 save                         # snapshot the process list
pm2 startup                      # prints a sudo line — run it once for boot survival

After pm2 startup, dhaara survives reboots. Any future pm2 stop / pm2 delete requires another pm2 save to persist.

Operating it

pm2 list                            # what's registered, what's running
pm2 logs dhaara                     # tail the bot's output
pm2 logs dhaara-weekly --lines 50   # tail any specific app
pm2 restart dhaara                  # bounce the bot after a code change
pm2 trigger dhaara-dashboard        # force a scheduled job to fire NOW (testing)
pm2 stop dhaara-streak              # disable an app — then `pm2 save`
pm2 delete dhaara-weekly            # unregister entirely — then `pm2 save`

Caveats and troubleshooting

stopped is the normal state for cron apps between fires. They wake up on schedule, run once, and go back to sleep. Only dhaara should show online 24/7.
PM2 uses local OS time. Match the timezone in your config.yaml. If the OS is UTC and the config says Asia/Kolkata, schedules above fire 5h30m later than you expect.
autorestart: false is mandatory for cron jobs. Without it, PM2 immediately re-runs the script on exit and "every Sunday 21:07" becomes "as fast as Python can finish." The shipped manifest already sets this correctly.
Bot silent on Telegram? pm2 logs dhaara — common causes: typo in token, two machines polling the same bot (409 Conflict: terminated by other getUpdates request), or model rate-limiting.
macOS notifications not appearing? Confirm osascript (Script Editor) is allowed to send notifications under System Settings → Notifications.

Linux users

The streak nudge uses macOS-only osascript. In ecosystem.config.js, swap the osascript line for:

'notify-send "Dhaara" "$MSG"'

(requires libnotify). Or drop the app entirely with pm2 delete dhaara-streak && pm2 save.

Choosing a model

Dhaara needs a model that supports system prompts + tool use + multi-turn chat. Anything modern from Anthropic, OpenAI, Google, Mistral, or Amazon Nova qualifies.

OpenRouter (recommended for getting started)

Set any model ID from openrouter.ai/models that supports tool calling. Examples:

Model	Strength
`anthropic/claude-sonnet-4.5`	Best quality, best reasoning
`anthropic/claude-haiku-4.5`	Fast and cheap
`openai/gpt-4o-mini`	Cheap general-purpose
`google/gemini-2.5-flash`	Very cheap, very fast
`mistralai/mistral-large`	EU-hosted option

AWS Bedrock

Model ID	~Price (in/out per 1M)	Notes
`us.amazon.nova-micro-v1:0`	$0.04 / $0.14	Cheapest; text only
`us.amazon.nova-lite-v1:0`	$0.06 / $0.24	Cheap + multimodal
`us.amazon.nova-pro-v1:0`	$0.80 / $3.20	Good default
`global.anthropic.claude-haiku-4-5-20251001-v1:0`	$1 / $5	Fast Anthropic
`global.anthropic.claude-sonnet-4-6`	$3 / $15	Best quality

Newer Anthropic models on Bedrock require a cross-region inference profile prefix (global., us., etc.). Models without tool-use support (titan-*, llama2-*, mistral-7b-*, etc.) will not work.

Bedrock IAM needs:

{
  "Version": "2012-10-17",
  "Statement": [{
    "Effect": "Allow",
    "Action": [
      "bedrock:InvokeModel",
      "bedrock:InvokeModelWithResponseStream",
      "bedrock:ListFoundationModels",
      "bedrock:GetFoundationModelAvailability"
    ],
    "Resource": "*"
  }]
}

Also enable your model in AWS Console → Bedrock → Model access.

How your data is stored

Journal data lives under data_dir (default ~/PAI/DhaaraData/). TELOS context lives alongside it under _telos/ (by default <data_dir>/../_telos) so every future PAI agent can share the same personal context.

~/PAI/
├── _telos/                     # shared across all PAI agents
│   ├── work.md                 # you edit these
│   └── personal.md
└── DhaaraData/
    └── journal/
        ├── 2026-04-14.md
        ├── 2026-04-15.md
        └── 2026-04-16.md

One file per day — that's it. Inside each daily file, four fixed sections hold everything:

# 2026-04-15 Journal

## [WORK]
- [10:32 AM] [WORK/meetings] Had standup with the team
- [2:15 PM] [WORK/coding] Finished the API refactor  *(mood: satisfied)*

## [PERSONAL]
- [9:00 AM] [PERSONAL/family] Had breakfast with family  *(mood: happy)*
- [11:00 PM] [PERSONAL/health] Couldn't sleep, felt anxious  *(mood: anxious)*

## [HABITS]
- [7:00 AM] [HABITS/exercise] Gym: 45 mins
- [10:00 PM] [HABITS/sleep] Sleep: 7 hours

## [FINANCE]
- [1:30 PM] [FINANCE/food] Spent ₹150 on lunch
- [6:00 PM] [FINANCE/groceries] Bought vegetables ₹300
- [8:00 PM] [FINANCE/rent] Paid rent ₹15,000

Four fixed categories: WORK, PERSONAL, HABITS, FINANCE. The agent picks exactly one for each entry.
Free-form subcategories: lowercase tags the agent invents as needed — meetings, coding, groceries, rent, hiring, gifts… whatever fits.
Inline metadata per bullet: [time], [CATEGORY/subcategory], and optional *(mood: ...)*. Each line stands on its own — great for future RAG.
Safe by construction: the journal store resolves every path inside data_dir and rejects anything that tries to escape.
Concurrent-safe: file-locked writes via filelock, so the agent can append, edit, and delete entries without races.

TELOS — give the agent context

TELOS is a personal-context framework from Daniel Miessler. Drop your work and life priorities into _telos/work.md and _telos/personal.md (by default in <data_dir>/../_telos/, shared across all PAI agents), and Dhaara will reason about your entries in that light. Example:

# Work TELOS
- Shipping Dhaara Phase 1 by end of April
- Learning Rust on the side
- Quarterly OKR: reduce infra bill 20%

The agent re-reads these on every turn, so edit freely — no restart needed.

Asking for TELOS insights

Ask the bot for reflection in plain English — "how am I tracking against my work goals this month?" / "is my spending aligned with my TELOS?" / "what should I course-correct on?" — and the agent calls a dedicated telos_insights tool that pulls the last N days of journal entries (capped at 90), bundles them with your TELOS files, and returns a structured response with three sections:

Stats — concrete numbers: entry counts, finance totals, habit streaks, mood patterns.
TELOS Alignment — which entries support each TELOS goal, which contradict, which goals see no action at all.
Recommendations — actionable course corrections based on the gap.

The tool also computes a data-quality flag before responding. If you have fewer than 3 days of entries in the requested window, the response leads with ⚠ INSUFFICIENT DATA. If coverage is below 30%, it leads with ⚠ LIMITED DATA. The agent is instructed to relay these warnings prominently rather than glossing over them — sparse journaling produces sparse insights, and pretending otherwise is worse than admitting it.

The optional telos_dir: config knob lets you point at a custom location (default is <data_dir>/../_telos/, shared across PAI agents).

Architecture

Telegram  ──►  Sarvam AI  ──►  AI Agent  ──►  Journal Store
(voice/text)   (STT + lang     (Bedrock or     (sandboxed
               detect +         OpenRouter,     markdown
               translation)     tool-use loop)  read/write)
                   ▲                                │
                   └──── reply translated back ─────┘

Key design choices:

Provider abstraction — src/ai/provider.py lets you plug in new AI backends without touching the agent loop.
Typed tool interface — the agent can only call a small, audited set of tools: record_entry, read_today, read_day, read_telos, telos_insights, list_entries, edit_entry, delete_entry. Every path is resolved inside data_dir; traversal attempts are rejected.
Local timezone — timestamps honour whatever IANA zone you set in config.yaml.
Stateless per conversation — /clear resets context; your daily markdown files are the long-term memory.

Commands

Command	Does
`/start`	Intro + health check
`/clear`	Reset conversation context (your journal data is untouched)

Everything else is an entry or a conversation.

Tools

Eleven standalone CLI scripts ship in scripts/ for setup, daily nudges, reflection, exploration, and export. They all read the same plain-markdown journal — no separate database, no schema migrations. Pure stdlib at runtime (PyYAML is optional, only needed for data_dir auto-discovery from config.yaml).

Setup (§ Quick start walks you through these on first run)

init.py — create data_dir/, journal/, _telos/ + seed example TELOS files
check_config.py — lint config.yaml (placeholders, missing keys, invalid timezones)

Daily nudges

today.py — single-day breakdown with finance subtotal and last-entry hint
streak.py — current streak, designed to live in your shell prompt

Reflection

weekly_summary.py — markdown weekly digest with week-over-week deltas
dashboard.py — combined "everything in one document" digest
activity_heatmap.py — GitHub-contribution-style calendar
mood_timeline.py — per-day mood heatmap
stats.py — period stats roll-up

Exploration & export

search.py — structured search with category/mood/date filters and surrounding context
tags.py — inventory the subcategories and moods you've used (counts + last-seen dates)
export_journal.py — CSV/JSON export with optional pivot aggregation

Export and pivots

Every entry already lives as plain markdown, so grep works. But if you want a structured view for spreadsheets or pandas, there's a standalone exporter:

# All entries to CSV (stdout)
python scripts/export_journal.py

# Last 30 days, finance only, to a file
python scripts/export_journal.py --since 30d --category FINANCE -o finance.csv

# JSON for a specific date range
python scripts/export_journal.py --from 2026-04-01 --to 2026-04-30 -f json

It parses every [time], [CATEGORY/subcategory], mood, and text into one row per bullet — no extra dependencies beyond the standard library (PyYAML is used only if you let it auto-discover data_dir from config.yaml).

For pivots instead of raw rows, add --group-by:

# How is my finance spending distributed across subcategories?
python scripts/export_journal.py --category FINANCE --group-by subcategory

# Cross-tab category × subcategory across the whole journal
python scripts/export_journal.py --group-by category,subcategory

# Mood frequency + first/last date you tagged each one
python scripts/export_journal.py --group-by mood -f json

Each pivot row carries the grouping columns plus count, sum_amount (₹ totals for FINANCE rows; empty otherwise), first_date, and last_date. Rows sort by spending desc when any group has amounts, else by count desc. Valid keys: category, subcategory, mood, date (any combination, comma-separated).

Quick stats

For a roll-up instead of raw rows, there's scripts/stats.py. It reuses the same parser and prints per-category counts, finance totals (with subcategory breakdown + top expenses), habit streaks, and mood distribution:

# Last 30 days, human-readable report
python scripts/stats.py --since 30d

# JSON for piping into a dashboard
python scripts/stats.py --since 30d -f json

# Habits-only view
python scripts/stats.py --category HABITS

Currency parsing is best-effort — it understands ₹, Rs, $, and bare numbers in FINANCE entries, plus k, lakh/lac, and cr/crore multipliers.

Combined dashboard

For an end-of-week or end-of-month review, scripts/dashboard.py weaves every visualization in the toolset into a single Markdown document — streak, snapshot, period summary, activity calendar, finance highlights, habits, mood timeline, and notable moments:

# Last 7 days, stdout
python scripts/dashboard.py

# Specific ISO week to a file
python scripts/dashboard.py --from 2026-04-13 --to 2026-04-19 \
    -o ~/PAI/DhaaraData/dashboards/2026-W16.md

# With week-over-week deltas
python scripts/dashboard.py --since 7d --compare-prev

# Last month
python scripts/dashboard.py --since 4w

Empty days/sections are gracefully omitted, so the report is always tight regardless of how much you journaled. Pure markdown — paste anywhere, archive as a long-term log, or pipe into a Telegram bot. With --compare-prev, a "Compared to the previous week" section lands between the period summary and the activity calendar, surfacing entry-count, spending, mood-drift, and top finance shifts vs. the same-length window before.

Today

The "what have I journaled this morning?" glance — a focused, single-day breakdown:

# What's in today's file?
python scripts/today.py

# Yesterday's recap
python scripts/today.py --date 2026-04-29

# Markdown for a daily review note
python scripts/today.py -f markdown

You get one section per category in WORK / PERSONAL / HABITS / FINANCE order, entries sorted by time, finance subtotal in the section header, and a "Moods today" line at the bottom.

Empty days render a "Nothing recorded yet" message that also surfaces the gap since your previous entry — Last entry: 13 days ago (2026-04-17 12:04 PM) — so the report turns into a nudge instead of a dead end. If there are no entries anywhere yet, it says so explicitly.

Streak nudge

A glanceable one-liner showing your current streak — designed to live in your shell prompt or status bar:

$ python scripts/streak.py
🔥 5-day streak

$ python scripts/streak.py --quiet         # just the number, for shell prompts
5

$ python scripts/streak.py --text          # full breakdown
🔥 Current streak: 5 days
   Longest streak: 12 days
   Last entry:     2026-04-30 (today)
   Total entries:  234 across 87 days

A habit you can see is a habit you keep — drop the short or --quiet form into your PROMPT_COMMAND, starship config, or tmux status bar, and you'll get a daily nudge for free.

Activity calendar

GitHub-contribution-graph-style heatmap that answers "am I journaling consistently?" — rows are weeks, columns are weekdays, cells are bucketed entry counts (▁▃▅▆▇). The header line surfaces longest streak, current streak, and best day at a glance.

# Last 12 weeks (default)
python scripts/activity_heatmap.py

# Last year
python scripts/activity_heatmap.py --since 1y

# Markdown table for embedding in a journal note
python scripts/activity_heatmap.py --since 4w -f markdown

Mood timeline

The agent already tags entries with optional moods. scripts/mood_timeline.py adds the missing time dimension — when did "anxious" appear? Was it a one-day spike or a sustained pattern?

# Last 30 days, ANSI heatmap (one row per mood, one column per day)
python scripts/mood_timeline.py

# Specific range, markdown for embedding in a journal note
python scripts/mood_timeline.py --from 2026-04-01 --to 2026-04-30 -f markdown

# JSON for a notebook / dashboard
python scripts/mood_timeline.py --since 6m -f json

Heatmap cells use bar characters (▁▂▃▄▅▆▇█) so density reads at a glance even on terminals without color support.

Search

Plain markdown is grep-able already, but scripts/search.py understands the entry structure — filter by category, mood, date range, regex, and case-sensitivity in one go:

# Substring search across the last 30 days
python scripts/search.py dhaara --since 30d

# All "anxious" moments in April
python scripts/search.py --mood anxious --from 2026-04-01 --to 2026-04-30

# Case-sensitive regex over WORK entries only
python scripts/search.py "API\b" --regex --match-case --category WORK

# JSON for piping
python scripts/search.py "claude" -f json

# Surrounding context: 2 entries before and after each hit
python scripts/search.py "rent" --context 2

Matches are highlighted in bold red on a TTY (use --color always|never to override). Exit code is 0 on hits, 1 on no matches — handy for shell scripting.

With --context N (or -C N), each match shows the N entries before and after it in chronological order. Adjacent windows merge automatically; non-adjacent blocks are separated by --. Actual matches are marked with ▸ so you can spot them within the context. JSON output shifts to a list-of-blocks shape with an is_match flag per entry.

Tag inventory

The agent invents subcategory names as you go (coding, food, subscriptions, family...). Over time you build up a personal classification system without ever planning one. scripts/tags.py surfaces what's there:

# Whole-journal inventory
python scripts/tags.py

# Just the last 90 days — what's my recent taxonomy?
python scripts/tags.py --since 90d

# JSON for piping
python scripts/tags.py -f json

Each subcategory shows count + last-seen date, sorted by count descending. Useful for spotting subcategories you've used once and never again, subcategories that have exploded and might want splitting, or moods you've stopped tagging. Same parser as the rest of the toolset — works against your existing journal with no setup.

Weekly summary

For a shareable markdown digest of a single week — entry counts, finance highlights, habit streaks, mood distribution, and the entries that carried a mood — there's scripts/weekly_summary.py:

# Last 7 days
python scripts/weekly_summary.py

# Specific ISO week (Mon-Sun)
python scripts/weekly_summary.py --week 2026-W17

# Add week-over-week deltas (entries, spending, mood drift, top finance shifts)
python scripts/weekly_summary.py --week 2026-W17 --compare-prev

# Save next to your daily files
python scripts/weekly_summary.py --week 2026-W17 -o ~/PAI/DhaaraData/weekly/2026-W17.md

Output is plain Markdown — paste it into your journal repo, share it with a friend, or schedule it as a Sunday-night cron. With --compare-prev, you also get a "Compared to the previous week" section with concrete numbers like Entries: 18 → 31 (+13, +72%) and food: ₹200 → ₹500 (+₹300).

Roadmap

Phase 1 ✅ — Journaling, voice, multilingual, expense extraction, category classification, edit/delete by line number
Phase 2 ✅ — Lexical search with date/category/mood filters and ±N context (scripts/search.py); daily and weekly markdown summaries (today.py, weekly_summary.py); CSV/JSON export with pivot aggregation (export_journal.py)
Phase 3 ✅ (mostly) — Mood timelines (mood_timeline.py), habit dashboards (activity_heatmap.py, dashboard.py), growth analysis (stats.py, week-over-week deltas)
Phase 4 🚧 — Vector / RAG retrieval ("what did I write semantically about career anxiety last quarter?"); cross-agent workflows in the wider PAI ecosystem

Contributing

Contributions are very welcome. Open an issue first for anything non-trivial so we can align on direction. Good first areas:

New AI providers (Anthropic direct, Gemini direct, local Ollama via src/ai/provider.py)
Phase 4 features: vector retrieval over past entries (the lexical foundation in search.py is ready to layer over)
Better expense parsing across currencies (the extract_amount regex in scripts/stats.py is the obvious starting point — currently handles ₹, Rs, $, plus k / lakh / lac / cr / crore multipliers)
More tests for the agent loop (src/ai/graph.py) — the helper modules are well-covered but the LangGraph state machine is not directly exercised yet
Linux notification path for dhaara-streak (currently macOS-only via osascript)

Running tests

pip install -r requirements-dev.txt
pytest

The suite covers the journal formatter and all eleven CLI scripts — 437 unit tests in ~0.25s. Pure stdlib fixtures, no network. CI runs them on every push and PR against Python 3.11, 3.12, and 3.13 (see .github/workflows/test.yml).

License

MIT — see LICENSE. Use it, fork it, make it yours.

Built in India. Named for the quiet stream of thoughts we all carry.

Name		Name	Last commit message	Last commit date
Latest commit History 36 Commits
.github/workflows		.github/workflows
scripts		scripts
src		src
tests		tests
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
config.example.yaml		config.example.yaml
ecosystem.config.js.example		ecosystem.config.js.example
pytest.ini		pytest.ini
requirements-dev.txt		requirements-dev.txt
requirements.txt		requirements.txt

Folders and files

Latest commit

History

Repository files navigation

Dhaara

Why Dhaara?

How it feels

Quick start

1. Clone and install

2. Configure

3. Verify your config (optional but recommended)

4. Initialize the data layout

5. Run

Running with PM2

One-time setup

Start and persist

Operating it

Caveats and troubleshooting

Linux users

Choosing a model

OpenRouter (recommended for getting started)

AWS Bedrock

How your data is stored

TELOS — give the agent context

Asking for TELOS insights

Architecture

Commands

Tools

Export and pivots

Quick stats

Combined dashboard

Today

Streak nudge

Activity calendar

Mood timeline

Search

Tag inventory

Weekly summary

Roadmap

Contributing

Running tests

License

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages