Tag: Experience Induction

Posted on

Episodic Memory: Experience Induction & Reuse

“Experience Induction & Reuse” in the context of Episodic Memory means that an AI agent doesn’t just store raw conversation logs—it extracts structured lessons from past successful (and failed) interactions, generalizes them into reusable patterns, and applies those patterns to new, similar situations.

CN: “经验归纳与复用”在情景记忆（Episodic Memory）的语境下，指的是 AI Agent 不只是存储原始对话日志，而是从过去的成功（和失败）交互中提取结构化的经验教训，归纳为可复用的模式，并在新的类似场景中应用这些模式。

Think of it like a seasoned detective vs a rookie:

CN: 用人话比喻，就像资深侦探 vs 新警察：

Rookie (无经验归纳)Seasoned Detective (有经验归纳)
ENReads every case file from scratch. Has to figure out the same patterns over and over.Has a “playbook” of past cases. Recognizes: “This looks like the Smith case from 2022—we found the killer by checking the alibi.”
CN每份卷宗都从头读起，同样的模式要反复摸索。有一本“过往案例手册”，一眼认出：“这像2022年的Smith案——当时我们通过核查不在场证明找到了凶手。”
ENTries the same failing interrogation tactic repeatedly.Learns from past mistakes: “Last time we pressured the suspect, they clammed up. This time we try a different approach.”
CN反复使用同样失败的审讯技巧。从过去错误中学习：“上次给嫌疑人施压，对方直接闭嘴了。这次换一种方法。”

Core Flow of Experience Induction & Reuse

The process consists of four key stages: ① Extraction (提取) ──► ② Storage (存储) ──► ③ Reflection (反思) ──► ④ Reuse (复用)

3.1 阶段1：Episode 提取 (Extraction)

aw conversation is transformed into a structured Episode record:

CN: 原始对话被转换为结构化的 Episode 记录：

class Episode(BaseModel):
    observation: str   # Context and setup - what happened / 上下文和场景——发生了什么
    thoughts: str      # Internal reasoning process - "I..." / 内部推理过程——“我...”
    action: str        # What was done and how - "I..." / 做了什么、怎么做——“我...”
    result: str        # Outcome and retrospective - "I..." / 结果和复盘——“我...”

This schema preserves the complete decision-making process, enabling the agent to learn not just what worked, but why it worked and how to replicate the success.

CN: 这个结构保留了完整的决策过程，让 Agent 不仅能学到什么有效，还能学到为什么有效以及如何复制成功。

3.2 阶段2：存储与检索 (Storage & Retrieval)

EN: Episodes are stored as vector embeddings in a vector database (e.g., Azure AI Search, Pinecone, Chroma). Retrieval is done by semantic similarity—not by recency.

CN: Episode 以向量嵌入的形式存储在向量数据库中（如 Azure AI Search、Pinecone、Chroma）。检索基于语义相似度，而不是时间先后。

EN: Key storage operations:

CN: 关键存储操作：

OperationMethodPurpose
EN: Store EpisodeEN: MemoryStoreManager.ainvoke()EN: Persist new episodes
CN: 存储EpisodeCN:CN: 持久化新的Episode
EN: Search SimilarEN: BaseStore.asearch()EN: Find relevant past experiences
CN: 相似检索CN:CN: 查找相关的过往经验
EN: Update EpisodeEN: enable_updates=TrueEN: Refine existing episodes
CN: 更新EpisodeCN:CN: 优化已有的Episode

3.3 阶段3：反思与归纳 (Reflection & Induction) ★核心★

EN: This is the heart of “experience induction.” The Reflection Module:

CN: 这是“经验归纳”的核心。反思模块（Reflection Module）：

  1. EN: Evaluates each episode for success or failure
    CN: 评估每个 Episode 是成功还是失败
  2. EN: Compares similar past episodes to identify generalizable patterns and principles
    CN: 比较相似的过往 Episode，识别可泛化的模式和原则
  3. EN: Distills reusable strategies rather than just replaying prior trajectories
    CN: 提炼可复用策略，而不仅仅是重放之前的轨迹

EN: For example:

CN: 例如：

成功Episode 1: 用 pandas chunking 处理 50GB CSV → 成功
成功Episode 2: 用 pandas chunking 处理 80GB Parquet → 成功
成功Episode 3: 用 pandas chunking 处理 120GB JSON → 成功

→ 反思归纳出模式: "处理大型数据集时，pandas + chunking 是可靠方案"
→ 可复用策略: "当数据量 > 10GB，使用 chunking 策略"

3.4 阶段4：经验复用 (Reuse)

EN: When a new task arrives:

CN: 当新任务到达时：

  1. EN: Agent queries episodic memory: “Have I done something similar before?”
    CN: Agent 查询情景记忆：“我以前做过类似的事吗？”
  2. EN: Relevant past episodes are retrieved by semantic similarity
    CN: 通过语义相似度检索相关的过往 Episode
  3. EN: Agent adapts its plan based on outcomes of past episodes
    CN: Agent 根据过往 Episode 的结果调整自己的计划
  4. EN: New success/failure becomes a new episode → loop continues
    CN: 新的成功/失败成为新的 Episode → 循环继续

Key Takeaways

要点 (Aspect)ENCN
Episodic Memory定义Episodic memory stores complete experiences: what happened, what was thought, what was done, and what resulted情景记忆存储完整经历：发生了什么、想了什么、做了什么、结果如何
经验归纳Reflection module compares similar episodes to extract generalizable patterns反思模块比较相似Episode，提取可泛化的模式
经验复用New tasks retrieve similar past episodes and adapt plans based on outcomes新任务检索相似过往Episode，根据结果调整计划
Episode结构observation → thoughts → action → resultobservation → thoughts → action → result
语义 vs 情景Semantic = facts (“what”); Episodic = experiences (“how”)语义 = 事实（”是什么”）；情景 = 经历（”怎么做”）
存储选型Vector database with semantic similarity + metadata filtering向量数据库 + 语义相似度 + 元数据过滤
核心价值Turns agent from stateless function into learning system that improves over time将Agent从无状态函数变成随时间持续改进的学习系统
LangMem工具Use create_memory_manager() with Episode schemas for extraction使用create_memory_manager()配合Episode结构进行提取
检索原则Retrieval works by relevance, not recency检索基于相关性，而非时间先后
Proudly powered by WordPress