Skip to content

Commit 849c951

Browse files
ViktarStarastsenkaViktarStarastsenka
committed
Updating the AI assistand guide
1 parent 1da2158 commit 849c951

File tree

1 file changed

+112
-106
lines changed

1 file changed

+112
-106
lines changed

‎src/uc/ai_assistant.md

Lines changed: 112 additions & 106 deletions
Original file line numberDiff line numberDiff line change
@@ -1,144 +1,150 @@
1-
Build a production-ready AI assistant that remembers conversations semantically and auto-cleans using Redis 8’s new features.
2-
3-
This smart AI assistant offers:
4-
- **Ephemeral Memory**: Messages expire automatically at different intervals
5-
- **Semantic Recall**: Search past chats by meaning, not just keywords
6-
- **Zero Maintenance**: No cleanup scripts or cron jobs needed
7-
- **Multi-User Support**: Memory isolated by user and session
8-
- **Hybrid Search**: Combine text and vector search for better results
9-
10-
**Note**: [Redis 8](https://hub.docker.com/_/redis/tags) is required for this tutorial Redis 8 since it introduces `HSETEX`, which lets you set TTL per hash field, ideal for ephemeral chat memory without complex expiration logic.
11-
12-
### Hierarchical Memory Structure
13-
Store chat sessions as Redis hashes. Each message is a field with its own TTL.
14-
15-
```redis:[run_confirmation=true] Upload Session Data
16-
// User-scoped sessions for isolation
17-
// Pattern: user:{user_id}:session:{session_id}
18-
HSETEX user:alice:session:morning EX 3600 FIELDS 1 msg:1717935301 "Good morning! What's my schedule today?"
19-
HSETEX user:alice:session:morning EX 1800 FIELDS 1 msg:1717935361 "Remind me about the team meeting at 2 PM"
20-
HSETEX user:alice:session:morning EX 900 FIELDS 1 msg:1717935420 "What's the weather forecast?"
21-
HSETEX user:alice:session:morning EX 300 FIELDS 1 msg:1717935480 "Thanks, that's all for now"
22-
23-
// Different user, same session pattern
24-
HSETEX user:bob:session:work EX 7200 FIELDS 1 msg:1717935500 "I need to prepare for the client presentation"
25-
HSETEX user:bob:session:work EX 3600 FIELDS 1 msg:1717935560 "What are the key points I should cover?"
1+
This intelligent AI assistant is designed to support real-world, multi-session use with Redis as its memory core.
2+
3+
What you get:
4+
- **Smart Memory**: Ephemeral context that expires automatically, long-term facts retained forever
5+
- **Semantic Search**: Recall relevant info by meaning using vector search
6+
- **Zero Maintenance**: Auto-expiring short-term memory without a need to track timestamps manually
7+
- **Multi-User**: Isolated memory per user
8+
- **Learning**: Assistant can "understand" each user better the more it's used
9+
10+
**Note**: Requires [Redis 8](https://hub.docker.com/_/redis/tags) for `HSETEX`, which adds per-field TTL for hashes, which is ideal for managing short-term memory with precision.
11+
12+
### Architecture Overview
13+
| Layer | Description |
14+
| ---------- | ---------- |
15+
| `Working Memory`| `Short-term chat context (ephemeral)` |
16+
| `Knowledge Base` | `Persistent facts, user preferences` |
17+
| `Vector Search` | `Unified semantic recall across both layers` |
18+
19+
### Working Memory (Ephemeral)
20+
Stores recent user messages with TTL based on importance. Automatically expires to prevent bloat.
21+
This uses `HSETEX`, a Redis 8 command that adds field-level expiration to hashes. It allows storing all temporary messages in a single key while managing TTLs per message, simplifying short-term memory management without needing multiple keys.
22+
23+
```redis:[run_confirmation=true] Recent Conversations with TTL Based on Importance.
24+
// Quick exchanges (5 min)
25+
HSETEX user:alice:session:001 EX 300 FIELDS 1 msg:001 "What's the weather?"
26+
// Session context (30 min)
27+
HSETEX user:alice:session:001 EX 1800 FIELDS 1 msg:002 "I need a dentist appointment"
28+
// Important decisions (2 hours)
29+
HSETEX user:alice:session:001 EX 7200 FIELDS 1 msg:003 "Book it for Tuesday 2 PM"
2630
```
2731

28-
### Memory Tiers for Different Lifetimes
29-
Control how long messages last depending on their importance.
30-
31-
```redis:[run_confirmation=true] Memory Tiers Strategy
32-
// Short-term (5 minutes) - Immediate context
33-
HSETEX user:alice:session:current EX 300 FIELDS 1 msg:1717935301 "Current conversation context"
34-
35-
// Medium-term (30 minutes) - Session memory
36-
HSETEX user:alice:session:current EX 1800 FIELDS 1 msg:1717935302 "Important session details"
37-
38-
// Long-term (2 hours) - Cross-session context
39-
HSETEX user:alice:session:current EX 7200 FIELDS 1 msg:1717935303 "Key user preferences and facts"
32+
### Knowledge Base (Persistent)
33+
Long-term memory: stores important facts, user preferences, and context across sessions. These never expire.
34+
`embedding` is a binary-encoded `FLOAT32[]` used for vector similarity that can be generated using sentence-transformers or similar libraries. Demo uses 8-dim vectors; production models typically use 128–1536 dimensions.
35+
36+
```redis:[run_confirmation=true] Important User Information That Never Expires.
37+
// User preferences - need vector fields for search
38+
HSET user:alice:knowledge:pref:001 user_id "alice" memory_type "knowledge" content "prefers mornings before 10 AM" importance 9 timestamp 1717935000 embedding "\x40\x20\x00\x00\x3f\x80\x00\x00\x40\x40\x00\x00\x3f\x40\x00\x00\x40\x60\x00\x00\x40\x00\x00\x00\x3f\x00\x00\x00\x40\x80\x00\x00"
39+
HSET user:alice:knowledge:pref:002 user_id "alice" memory_type "knowledge" content "likes detailed explanations" importance 8 timestamp 1717935000 embedding "\x3f\x40\x00\x00\x40\x60\x00\x00\x40\x00\x00\x00\x40\x20\x00\x00\x3f\x80\x00\x00\x40\x40\x00\x00\x40\x80\x00\x00\x3f\x00\x00\x00"
40+
// Personal facts
41+
HSET user:alice:knowledge:personal:001 user_id "alice" memory_type "knowledge" content "allergic to shellfish" importance 10 timestamp 1717935000 embedding "\x40\x00\x00\x00\x40\x20\x00\x00\x3f\x80\x00\x00\x40\x40\x00\x00\x40\x60\x00\x00\x3f\x40\x00\x00\x40\x80\x00\x00\x3f\x00\x00\x00"
42+
HSET user:alice:knowledge:personal:002 user_id "alice" memory_type "knowledge" content "golden retriever named Max" importance 7 timestamp 1717935000 embedding "\x3f\x80\x00\x00\x40\x40\x00\x00\x40\x00\x00\x00\x3f\x40\x00\x00\x40\x80\x00\x00\x40\x20\x00\x00\x3f\x00\x00\x00\x40\x60\x00\x00"
43+
// Work context
44+
HSET user:alice:knowledge:work:001 user_id "alice" memory_type "knowledge" content "Senior PM at TechCorp" importance 8 timestamp 1717935000 embedding "\x40\x40\x00\x00\x3f\x00\x00\x00\x40\x80\x00\x00\x40\x20\x00\x00\x3f\x80\x00\x00\x40\x60\x00\x00\x40\x00\x00\x00\x3f\x40\x00\x00"
45+
HSET user:alice:knowledge:work:002 user_id "alice" memory_type "knowledge" content "leading Project Apollo" importance 9 timestamp 1717935000 embedding "\x40\x60\x00\x00\x40\x80\x00\x00\x3f\x40\x00\x00\x40\x00\x00\x00\x40\x20\x00\x00\x3f\x80\x00\x00\x40\x40\x00\x00\x3f\x00\x00\x00"
4046
```
4147

42-
### Check Current Session Memory
43-
No manual cleanup needed; expired messages vanish automatically.
44-
45-
```redis:[run_confirmation=true] Monitor Session State Over Time
46-
// After a few minutes, run this command to see what's left.
47-
HGETALL user:alice:session:morning
48-
```
49-
50-
### Vector Search Setup for Semantic Recall
51-
Create an index to store messages as vectors for semantic search.
48+
### Vector Search: Semantic Memory Recall
49+
Unify working + knowledge memory into a vector index for semantically meaningful search.
5250

5351
```redis:[run_confirmation=true] Create a Vector Index
54-
FT.CREATE idx:ai_memory
52+
FT.CREATE idx:memory
5553
ON HASH
56-
PREFIX 1 memory:
54+
PREFIX 2 vmemory: user:
5755
SCHEMA
58-
user_id TAG SORTABLE
59-
session_id TAG SORTABLE
60-
message TEXT
61-
context TEXT
62-
timestamp NUMERIC SORTABLE
56+
user_id TAG
57+
memory_type TAG
58+
content TEXT
59+
importance NUMERIC
60+
timestamp NUMERIC
6361
embedding VECTOR HNSW 6
6462
TYPE FLOAT32
65-
DIM 8 // DIM = embedding size, DIM 8 is just for demo purposes. In real use, embeddings are usually 128–1536
63+
DIM 8 // DIM = embedding size; 8 used here for simplicity — in production, use 128 to 1536
6664
DISTANCE_METRIC COSINE // COSINE = measures semantic closeness
6765
```
6866

69-
Add sample vectorized messages (embedding dims are demo-sized):
67+
### Add Indexed Memory
68+
Populate the vector index with memory items from both ephemeral and persistent layers.
7069

7170
```redis:[run_confirmation=true] Add entries for the chatbot
72-
HSET memory:alice:1 user_id "alice" session_id "morning" message "I have a dentist appointment at 3 PM today" context "healthcare scheduling appointment" timestamp 1717935301 embedding "\x3f\x00\x00\x00\x40\x00\x00\x00\x40\x40\x00\x00\x40\x80\x00\x00\x40\x00\x00\x00\x40\x00\x00\x00"
73-
HSET memory:alice:2 user_id "alice" session_id "morning" message "Remind me to water the plants in my office" context "task reminder plants office" timestamp 1717935361 embedding "\x40\x00\x00\x00\x40\x00\x00\x00\x40\x00\x00\x00\x3f\x00\x00\x00\x3f\x00\x00\x00\x40\x00\x00\x00"
74-
HSET memory:alice:3 user_id "alice" session_id "work" message "Schedule a meeting with the engineering team" context "work scheduling meeting team" timestamp 1717935420 embedding "\x40\x40\x00\x00\x40\x00\x00\x00\x40\x00\x00\x00\x40\x00\x00\x00\x3f\x00\x00\x00\x3f\x00\x00\x00"
75-
HSET memory:bob:1 user_id "bob" session_id "work" message "I need to review the quarterly sales report" context "business analysis quarterly report" timestamp 1717935480 embedding "\x40\x40\x00\x00\x40\x00\x00\x00\x3f\x00\x00\x00\x40\x00\x00\x00\x40\x00\x00\x00\x00\x00\x00\x00"
71+
// Working memory (expires from Redis, stays in search until rebuild)
72+
HSET vmemory:alice:001 user_id "alice" memory_type "working" content "Book dentist for Tuesday 2 PM" importance 8 timestamp 1717935310 embedding "\x3f\x80\x00\x00\x40\x00\x00\x00\x40\x40\x00\x00\x40\x80\x00\x00\x3f\x00\x00\x00\x40\x20\x00\x00\x40\x60\x00\x00\x3f\x40\x00\x00"
73+
74+
// Knowledge base (persistent)
75+
HSET vmemory:alice:kb:001 user_id "alice" memory_type "knowledge" content "allergic to shellfish" importance 10 timestamp 1717935000 embedding "\x40\x00\x00\x00\x40\x20\x00\x00\x3f\x80\x00\x00\x40\x40\x00\x00\x40\x60\x00\x00\x3f\x40\x00\x00\x40\x80\x00\x00\x3f\x00\x00\x00"
7676
```
7777

78-
### Let Chatbot Think – Semantic Search with Vectors
79-
When a user says something new, find all related past conversations across your entire system based on semantic meaning.
78+
### Full Memory Search
79+
Search across all memories to recall relevant data.
8080

8181
```redis:[run_confirmation=false] Find Top 5 Related Messages By Meaning
82-
FT.SEARCH idx:ai_memory
83-
"*=>[KNN 5 @embedding $vec AS score]"
84-
PARAMS 2 vec "\x00\x00@@\x00\x00\x80@\x00\x00\x00@\x00\x00\x80?\x00\x00@@\x00\x00\x00@"
85-
RETURN 4 user_id message context timestamp
86-
SORTBY score ASC
87-
DIALECT 2
82+
FT.SEARCH idx:memory
83+
"(@user_id:{alice}) => [KNN 5 @embedding $vec AS score]"
84+
PARAMS 2 vec "\x40\x00\x00\x00\x40\x20\x00\x00\x3f\x80\x00\x00\x40\x40\x00\x00\x40\x60\x00\x00\x3f\x40\x00\x00\x40\x80\x00\x00\x3f\x00\x00\x00"
85+
RETURN 4 content memory_type importance score
86+
SORTBY score ASC
87+
DIALECT 2
8888
```
8989

90-
Now your assistant “remembers” things it’s heard before - by meaning.
91-
92-
### User-Scoped Semantic Search
93-
Your AI should only recall memories from the specific user it's talking to, not leak information between users.
90+
### Session-Only Search
91+
Limit results to current conversation context (ephemeral memory).
9492

95-
```redis:[run_confirmation=false] Find Similar Memories For Specific User Only
96-
FT.SEARCH idx:ai_memory
97-
"(@user_id:{alice}) => [KNN 3 @embedding $query_vec AS vector_score]"
98-
PARAMS 2 query_vec "\x00\x00@@\x00\x00\x80@\x00\x00\x00@\x00\x00\x80?\x00\x00@@\x00\x00\x00@"
99-
RETURN 5 user_id message context vector_score session_id
100-
SORTBY vector_score ASC
93+
```redis:[run_confirmation=false] Session-Only Search
94+
FT.SEARCH idx:memory
95+
"(@user_id:{alice} @memory_type:{working}) => [KNN 5 @embedding $vec AS score]"
96+
PARAMS 2 vec "\x3f\x80\x00\x00\x40\x40\x00\x00\x40\x00\x00\x00\x3f\x40\x00\x00\x40\x80\x00\x00\x40\x20\x00\x00\x3f\x00\x00\x00\x40\x60\x00\x00"
97+
RETURN 4 content score timestamp memory_type
98+
SORTBY score ASC
10199
DIALECT 2
102100
```
103101

104-
### Time-Bounded Semantic Search
105-
When users ask about "recent" things, limit your search to a specific time window while still using semantic matching.
102+
### Knowledge-Only Search
103+
Focus only on persistent memory: facts, preferences, decisions.
106104

107-
```redis:[run_confirmation=false] Find recent similar memories (last 24 hours)
108-
FT.SEARCH idx:ai_memory
109-
"(@timestamp:[1717849200 +inf]) => [KNN 3 @embedding $query_vec AS vector_score]"
110-
PARAMS 2 query_vec "\x00\x00@@\x00\x00\x80@\x00\x00\x00@\x00\x00\x80?\x00\x00@@\x00\x00\x00@"
111-
RETURN 6 message timestamp vector_score
112-
SORTBY timestamp DESC
105+
```redis:[run_confirmation=false] Knowledge-Only Search
106+
FT.SEARCH idx:memory
107+
"(@user_id:{alice} @memory_type:{knowledge}) => [KNN 8 @embedding $vec AS score]"
108+
PARAMS 2 vec "\x40\x40\x00\x00\x3f\x00\x00\x00\x40\x80\x00\x00\x40\x20\x00\x00\x3f\x80\x00\x00\x40\x60\x00\x00\x40\x00\x00\x00\x3f\x40\x00\x00"
109+
RETURN 4 content importance score timestamp
110+
SORTBY score ASC
113111
DIALECT 2
114112
```
115113

116-
### Session-Specific Recall
117-
When users refer to something from "earlier in our conversation," search only within the current session context.
114+
### Monitoring Memory State
115+
Use these queries to inspect what’s stored in memory.
118116

119-
```
120-
FT.SEARCH idx:ai_memory
121-
"(@user_id:{alice} @session_id:{morning}) => [KNN 10 @embedding $query_vec AS vector_score]"
122-
PARAMS 2 query_vec "\x00\x00@@\x00\x00\x80@\x00\x00\x00@\x00\x00\x80?\x00\x00@@\x00\x00\x00@"
123-
RETURN 4 message context vector_score timestamp
124-
SORTBY vector_score
125-
DIALECT 2
117+
```redis:[run_confirmation=false] Check Memory State
118+
// Check active session memory
119+
HGETALL user:alice:knowledge:pref:001 // Example for one preference item
120+
121+
// View user knowledge
122+
HGETALL user:alice:knowledge:pref:001
123+
124+
// Search user's memories
125+
FT.SEARCH idx:memory
126+
"@user_id:{alice}"
127+
RETURN 3 content memory_type importance
126128
```
127129

128-
### Want to check what's still in memory?
130+
### Data Cleanup
129131

130-
Only unexpired fields remain.
132+
For privacy compliance, delete all user-related keys.
131133

132-
```redis:[run_confirmation=false] Check Sessions
133-
HGETALL memory:alice:1
134-
HGETALL memory:alice:2
135-
HGETALL memory:alice:3
136-
HGETALL memory:bob:1
134+
```redis:[run_confirmation=true] Complete user removal
135+
DEL user:alice:knowledge:pref:001
136+
DEL user:alice:knowledge:pref:002
137+
DEL user:alice:knowledge:personal:001
138+
DEL user:alice:knowledge:personal:002
139+
DEL user:alice:knowledge:work:001
140+
DEL user:alice:knowledge:work:002
141+
DEL vmemory:alice:001
142+
DEL vmemory:alice:kb:001
137143
```
138144

139145
### Next Steps
140146
Now that your assistant has memory and meaning, you can:
141-
- Combine Redis Vector Search with LLMs (RAG)
142-
- Use OpenAI or sentence-transformers for embeddings
143-
- Add fallback persistent storage with Redis Flex
144-
- Manage users with ACLs, quotas, and keyspace notifications
147+
- Combine with RAG Pipelines
148+
- Use sentence-transformers to generate high-dimensional vectors
149+
- Add [Redis Flex](https://redis.io/solutions/flex/?utm_source=redisinsight&utm_medium=app&utm_campaign=tutorials) for fallback persistence
150+
- Use Redis ACLs to isolate users, enforce quotas, and monitor usage

0 commit comments

Comments
 (0)