my-project-public

repository

loading code, commits, and activity

repositories

loading repo index

#1	# Getting Started With Clawd Memory
#2
#3	This guide sets up the Clawd agent memory layer locally.
#4
#5	## 1. Install
#6
#7	From the source tree:
#8
#9	```bash
#10	cd /Users/8bit/bots/Cladwbot-solana/solana-clawd/MemeBRain
#11	python3 -m pip install -e ".[all]"
#12	```
#13
#14	If you only need the CLI from the checked-out repo, you can run the module directly without installing:
#15
#16	```bash
#17	python3 -m mnemosyne.clawd_brain --help
#18	```
#19
#20	Package compatibility note: the Python package is still named `mnemosyne-memory`, and the module is still `mnemosyne`. The Clawd-facing command is `clawd-brain`.
#21
#22	## 2. Initialize The Clawd Bank
#23
#24	```bash
#25	python3 -m mnemosyne.clawd_brain init
#26	```
#27
#28	Expected output includes:
#29
#30	- `bank`: `clawd`
#31	- `vault`: the markdown vault path
#32	- `index_db`: the Clawd vault index database
#33	- `memory`: engine statistics
#34
#35	Override the vault location when needed:
#36
#37	```bash
#38	export CLAWD_BRAIN_VAULT="$PWD/vault"
#39	python3 -m mnemosyne.clawd_brain init
#40	```
#41
#42	## 3. Remember Durable Agent Context
#43
#44	```bash
#45	python3 -m mnemosyne.clawd_brain remember \
#46	"Clawd Agent Contract" \
#47	"Recall before planning. Remember durable decisions, preferences, protocol research, and risk findings. Never store secrets." \
#48	--kind agent \
#49	--tag clawd \
#50	--tag agent \
#51	--importance 0.9
#52	```
#53
#54	Use short titles and direct content. The content should be useful if another agent recalls it days later.
#55
#56	## 4. Recall Context
#57
#58	```bash
#59	python3 -m mnemosyne.clawd_brain recall "agent memory rules" --top-k 8
#60	```
#61
#62	Recall returns matching memory records and vault notes. Agents should run recall before making decisions that depend on prior context.
#63
#64	## 5. Archive Research
#65
#66	Archive a URL:
#67
#68	```bash
#69	python3 -m mnemosyne.clawd_brain research "https://docs.jup.ag/" --tag solana --tag jupiter
#70	```
#71
#72	Queue a topic:
#73
#74	```bash
#75	python3 -m mnemosyne.clawd_brain research "BONK perpetual venue risk" --tag perp --tag risk
#76	```
#77
#78	The research command creates a durable vault note and indexes it for later recall.
#79
#80	## 6. Ingest OODA Ticks
#81
#82	```bash
#83	python3 -m mnemosyne.clawd_brain ingest-ooda --journal ../ooda/journal/ticks.jsonl --limit 100
#84	```
#85
#86	Use this when the OODA loop has produced operational observations that should become durable Clawd memory.
#87
#88	## 7. Check Status
#89
#90	```bash
#91	python3 -m mnemosyne.clawd_brain status
#92	```
#93
#94	Status shows the active bank, vault path, vault note counts, link count, note counts by kind, and underlying memory stats.
#95
#96	## Recommended Agent Workflow
#97
#98	1. Recall relevant context for the user/task.
#99	2. Do the work.
#100	3. Remember decisions, preferences, risk findings, and deployment facts.
#101	4. Archive URLs or research topics that should be reusable.
#102	5. Consolidate periodically through the Hermes/MCP `mnemosyne_sleep` tool or the lower-level engine.
#103
#104	## What Not To Store
#105
#106	- API keys, private keys, seed phrases, tokens, passwords, or session cookies.
#107	- Raw logs that contain secrets or customer data.
#108	- Large command output that does not create future context.
#109	- Temporary thoughts that belong in scratchpad only.
#110
#111	## Next
#112
#113	- Read [Architecture](architecture.md) for the data model.
#114	- Read [Configuration](configuration.md) for environment variables.
#115	- Read [API Reference](api-reference.md) for Python usage.
#116

z6Mkq5mY3JWtxoxUobWcfNHm7AkRubgSWEZTkBVqZXJviFZ5/my-project-public