Document templates

Build templates with placeholders that mechanically substitute from matter data, then have the AI fill any free-form narrative gaps. Live at /admin/legal/templates.

Placeholder vocabulary

{{matter.title}}             matter title
{{matter.number}}            matter number
{{matter.type}}              type (litigation / m_and_a / ...)
{{matter.court}}             court
{{matter.docket}}            docket number
{{matter.jurisdiction}}      jurisdiction
{{matter.description}}       free-text description
{{matter.opened}}            ISO date matter opened
{{matter.billing}}           billing model (hourly / flat / contingency)
{{matter.flat_fee_usd}}      if flat-fee
{{client.name}}              display name
{{client.first_name}}        for individuals
{{client.last_name}}         for individuals
{{client.organization_name}} for orgs
{{client.email}}             email
{{client.phone}}             phone
{{client.address}}           street
{{client.city}}              city
{{client.state}}             state
{{client.postal_code}}       ZIP
{{parties.0.name}}           first party's name
{{parties.0.role}}           first party's role
{{today}}                    YYYY-MM-DD
{{today_long}}               "May 24, 2026"

Unresolved placeholders render as [[matter.field]] so the user sees what’s missing.

Narrative blocks

For free-form sections, write:

{{narrative: describe the scope of engagement based on matter type and any specifics}}

After mechanical substitution, the LLM is asked to replace every {{narrative: ...}} block with prose that matches the surrounding register. Original draft stays byte-identical outside those blocks.

LLM system prompt

Templates can carry a custom llm_system_prompt that overlays the default narrative-pass system message. Useful for “always cite California law” or “match our firm’s formal-counsel register” templates.

Generating

From the template card, click Generate for matter → pick a matter from the dropdown → Generate. Output renders in a read-only block with a Copy button. Stats above the output show how many placeholders filled and how many narrative blocks resolved.

Example — engagement letter

Dear {{client.name}},

This letter confirms our representation in {{matter.title}}
(matter no. {{matter.number}}).

{{narrative: describe scope of engagement based on matter type and
any specifics}}

Our fees will be billed {{matter.billing}}.

Very truly yours,

Click Generate → engagement letter pre-filled with client + matter + billing model, with a fully-written scope paragraph the AI synthesized from matter data.