Skip to content

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

模板优化 (#3242)

fujiajie6664847ffa378c5
diff --git a/benchmark/locomo/vikingbot/run_eval.py b/benchmark/locomo/vikingbot/run_eval.py
index 1046a10b1..79cf54c7c 100644
--- a/benchmark/locomo/vikingbot/run_eval.py
+++ b/benchmark/locomo/vikingbot/run_eval.py
@@ -202,7 +202,7 @@ def load_locomo_qa(
"question_id": question_id,
"question_index": question_index,
"question": qa["question"],
- "answer": qa["answer"],
+ "answer": qa.get("answer", qa.get("adversarial_answer", "")),
"category": qa.get("category", ""),
"evidence": evidence_list,
"evidence_text": get_evidence_text(evidence_list, sample),
@@ -224,7 +224,7 @@ def load_locomo_qa(
"question_id": question_id,
"question_index": q_idx,
"question": qa["question"],
- "answer": qa["answer"],
+ "answer": qa.get("answer", qa.get("adversarial_answer", "")),
"category": qa.get("category", ""),
"evidence": evidence_list,
"evidence_text": get_evidence_text(evidence_list, sample),
diff --git a/openviking/prompts/templates/memory/preferences.yaml b/openviking/prompts/templates/memory/preferences.yaml
index ca65a9e3c..4da91bb44 100644
--- a/openviking/prompts/templates/memory/preferences.yaml
+++ b/openviking/prompts/templates/memory/preferences.yaml
@@ -37,12 +37,15 @@ fields:
{% if language == 'en' %}Use lowercase with underscores, max 3 words.{% else %}Keep the topic concise and natural in {{ language }}.{% endif %}
Preference topic used to uniquely identify this preference memory.
Should be a semantic topic description such as "Python code style", "Communication style", "Food preference", "Commute preference", etc.
- Different preference topics should be stored as separate memories, do not mix unrelated preferences.
+ Store unrelated preferences separately, but merge overlapping or synonymous topics into one stable topic.
merge_op: immutable
- name: content
type: string
description: |
Preference content in Markdown format describing "what the user prefers/is accustomed to".
- Example: "User has shown clear preferences for Python code style in multiple conversations: dislikes using type hints, considers them redundant; requires concise function comments, limited to 1-2 lines; prefers direct implementation, avoids excessive fallbacks and over-engineering."
+ Write one concise bullet per distinct, reusable preference; omit duplicates, events, dates, plans, and supporting evidence.
+ Example:
+ - Prefers concise written updates with decisions and next actions.
+ - Chooses lightweight tools when they cover the workflow.
merge_op: patch
diff --git a/openviking/prompts/templates/memory/profile.yaml b/openviking/prompts/templates/memory/profile.yaml
index 9d3c67ef7..561f8ddf7 100644
--- a/openviking/prompts/templates/memory/profile.yaml
+++ b/openviking/prompts/templates/memory/profile.yaml
@@ -1,8 +1,8 @@
memory_type: profile
description: |
# Task Objective
- User profile memory - captures "who the user is" as a person.
- Extract relatively stable personal attributes that define the user's identity, work style, and preferences.
+ User profile memory - captures an identity summary of who the user is as a person.
+ Extract relatively stable personal attributes that define the user's identity at a high level.
Include: profession, experience level, technical background, communication style, work habits, etc.
Do NOT include transient conversation content or temporary mood states.
@@ -22,6 +22,7 @@ fields:
type: string
description: |
User profile content in Markdown format describing "who the user is".
+ Use Markdown heading(s) with only the person's name, then bullets only; do not write paragraph bodies.
Includes relatively stable personal attributes: profession, experience, tech stack, communication style, etc.
Only record objective statuses, do not record events or similar information.
[IMPORTANT] For changeable statuses, must include the last updated time in the format: (as of 2023-06-09)