It has been around 1.5 years since we started working on an open-source project named Glific that would empower communication and help NGOs deliver their programs digitally and manage communication with their beneficiaries at scale.
One of the core features of Glific is the integration of flow editor which helps in creating custom step by step flow to automate conversations while also saving user’s responses like name, age, gender, etc as contact fields, which helps NGOs greatly to deliver their programs to a significantly large number of beneficiaries.
Digital Green and their vision:
With our newest partnership with DigitalGreen aims at supporting smallholder Chilli farmers in Andhra Pradesh by helping them with preventive steps from Leaf Curl, a plantation disease, and curation steps if their crop is already manifested with the disease during the crop cycle.
These advisory messages need to be sent twice a day on a defined time slot and as the crop stage is determined by a number of days since sowing, preventive and curative steps for the crop varies with the stage crop is in, therefore it requires some computation based on the day the farmer joined the program, initial crop stage, the climatic condition of nearby area and condition of the crop before sending message.
Through our calls and discussion, we listed down their requirements:
- Ability to capture the day farmer enrolled and the current stage of the crop
- Computing number days since enrolled date and determining crop stage
- Sending preventive/ curative steps based on crop stage and status of crop if it is infected
- Sending Timely reminder at defined time slots
As we started creating their flows, we figured out that we need to tweak the flow editor a bit in order to support DigitalGreen’s need as in FlowEditor replies were sent based on the user’s response but now we need automated replies based on the crop stage and climatic condition of the nearby area on defined time slots.
Revising automated conversation
Capturing the datapoints: We researched and planned possible scenarios on how we can update Flow editor, and decided to use EEx support in Flow editor, allowing embedding Elixir code inside Flow Editor. With EEx we were able to capture the day the farmer enrolled and the crop stage.
With the enrolled date and crop stage set, next, we need to compute the number of days since farmers enrolled and categorize them in different stages based on days.
Categorizing Farmers based on different stages: For categorizing, we added support in Flow Editor to call an Elixir function, which was an extension to Flow Editor’s inbuilt function Call-Webhook.
With Call-a-Function we can call a particular elixir function from our codebase with contact fields like enrolled day, crop stage, etc as function arguments to compute the total days of the crop to update crop stage and categorize.
Based on different crop stages we add farmers to different collections in Glific.
Sending Timely Advisory steps: For sending preventive/ curative steps on a timely basis, we used Glific’s other feature: Triggers which lets you send recurring messages at a scheduled time.
For the last piece of the puzzle, we added a dynamic wait for a response to wait till the next time slot to send the message when in a flow in case there is no response from the user’s end.
As speech-based conversations have several potential benefits for farmers, we will be integrating audio to text translation with help from Navana Tech to help farmers to communicate more easily with bot
With Glific, we try to solve most of the NGOs needs right out of the box and are always happy to hear back from the community and their needs, to make Glific a better fit.
[…] Digital Green: Chatbot based farming advisor – Akhilesh Negi […]