Hizal Search#

Search before you build. Don't rediscover what the team already decided.

Semantic Search#

Start with 2-3 broad searches using different phrasings:

text

hizal__search_context(query="<key concept from the task>")
hizal__search_context(query="<ticket id or feature name>")
hizal__search_context(query="<related subsystem or endpoint>")

Narrow by Scope#

text

# Project-specific knowledge and conventions
hizal__search_context(query="<concept>", scope="PROJECT", project_id="<id>")

# Prior agent memory / investigation notes
hizal__search_context(query="<concept>", scope="AGENT", chunk_type="MEMORY")

# Org-wide principles and standards
hizal__search_context(query="<concept>", scope="ORG")

Filter by Chunk Type#

text

hizal__search_context(query="<concept>", chunk_type="KNOWLEDGE")
hizal__search_context(query="<concept>", chunk_type="CONVENTION")
hizal__search_context(query="<concept>", chunk_type="MEMORY")

Exact Lookup by query_key#

When you know the exact key:

text

hizal__read_context(query_key="<exact-query-key>", project_id="<id>")

Read a Specific Chunk#

text

hizal__read_context(id="<chunk-uuid>")

Synthesize Related Context#

Pull related chunks for synthesis (read-only, never delete source chunks):

text

hizal__compact_context(query="<concept>")
hizal__compact_context(query="<concept>", scope="PROJECT", project_id="<id>")

Version History#

Inspect how a chunk evolved:

text

hizal__get_context_versions(id="<chunk-uuid>")

Search Strategy#

1. Broad first — 2-3 searches with different phrasings
2. Read the results — chunks contain architecture decisions, conventions, and prior work
3. Narrow by scope if broad search returns too much noise
4. Use query_key if you already know the key
5. If an AGENT memory chunk is broadly useful — promote it with write_knowledge or write_convention

See hizal-write skill for writing back to Hizal.