Support Non-AI Templating

[charlesyu108]

We currently support synthesizing/compiling email using an LLM as a template engine. We should also support simple use cases where you might not want to use AI. Engines like Jinja & Nunchucks already exist for this.

Idea: Build an abstractions for a Renderer: AIRenderer or JinjaRenderer. Switching which one to use will be based on the flag --model

[monikkacha]

Hey @charlesyu108

I just checked recent code for local LLM and it looks cool, starting to work on Nunjucks non-AI feature .

[monikkacha]

Hey @charlesyu108

Facing an issue due to running mailmerge directly from the project, setup operation works as expected but the compose operation gives an error about not passing the contact option though it's provided in the command.

Please let me know how do you handle mailmerge operations during development, it might sound silly but working with CLI project first time 😅

[charlesyu108]

Hey @charlesyu108

Facing an issue due to running mailmerge directly from the project, setup operation works as expected but the compose operation gives an error about not passing the contact option though it's provided in the command.

Please let me know how do you handle mailmerge operations during development, it might sound silly but working with CLI project first time 😅

Ah, gotcha. If you are running with npm start you'll need to pass -- in between the args otherwise, npm treats the flags as args to npm, so: npm start -- compose --contacts ./example/sample-contacts.csv ./examples/outreach-template.md

[monikkacha]

Thanks for the help man, Nunjucks engine added and raised PR #11 , let me know if any changes required .

[charlesyu108]

Thanks for the help man, Nunjucks engine added and raised PR #11 , let me know if any changes required .

Thanks! Super fast. 1 small change requested but otherwise looks great!