Packages
fnord
0.9.21
0.9.40
0.9.39
0.9.38
0.9.37
0.9.36
0.9.35
0.9.34
0.9.33
0.9.32
0.9.31
0.9.30
0.9.29
0.9.28
0.9.27
0.9.26
0.9.25
0.9.24
0.9.23
0.9.22
0.9.21
0.9.20
0.9.19
0.9.18
0.9.17
0.9.16
0.9.15
0.9.14
0.9.13
0.9.12
0.9.11
0.9.10
0.9.9
0.9.8
0.9.7
0.9.6
0.9.5
0.9.4
0.9.3
0.9.2
0.9.1
0.9.0
0.8.99
0.8.98
0.8.97
0.8.96
0.8.95
0.8.94
0.8.93
0.8.92
0.8.91
0.8.90
0.8.89
0.8.88
0.8.87
0.8.86
0.8.85
0.8.84
0.8.83
0.8.82
0.8.81
0.8.80
0.8.79
0.8.78
0.8.77
0.8.76
0.8.75
0.8.74
0.8.73
0.8.72
0.8.71
0.8.70
0.8.69
0.8.68
0.8.67
0.8.66
0.8.65
0.8.64
0.8.63
0.8.62
0.8.61
0.8.60
0.8.59
0.8.58
0.8.57
0.8.56
0.8.55
0.8.54
0.8.53
0.8.52
0.8.51
0.8.50
0.8.49
0.8.48
0.8.47
0.8.46
0.8.45
0.8.44
0.8.43
0.8.42
0.8.41
0.8.40
0.8.39
0.8.38
0.8.37
0.8.36
0.8.35
0.8.34
0.8.33
0.8.32
0.8.31
0.8.30
0.8.29
0.8.27
0.8.26
0.8.25
0.8.24
0.8.23
0.8.22
0.8.21
0.8.20
0.8.19
0.8.18
0.8.17
0.8.16
0.8.15
0.8.14
0.8.13
0.8.12
0.8.11
0.8.1
0.8.0
0.7.24
0.7.23
0.7.22
0.7.21
0.7.20
0.7.19
0.7.18
0.7.17
0.7.16
0.7.15
0.7.14
0.7.13
0.7.12
0.7.11
0.7.10
0.7.9
0.7.8
0.7.7
0.7.6
0.7.5
0.7.3
0.7.2
0.7.1
0.7.0
0.6.9
0.6.8
0.6.7
0.6.6
0.6.5
0.6.4
0.6.3
0.6.1
0.6.0
0.5.9
0.5.8
0.5.7
0.5.6
0.5.5
0.5.4
0.5.3
0.5.2
0.5.1
0.5.0
0.4.44
0.4.43
0.4.42
0.4.41
0.4.40
0.4.39
0.4.38
0.4.37
0.4.36
0.4.35
0.4.34
0.4.33
0.4.32
0.4.30
0.4.29
0.4.28
0.4.27
0.4.26
0.4.25
0.4.24
0.4.23
0.4.22
0.4.21
0.4.20
0.4.19
0.4.18
0.4.17
0.4.16
0.4.15
0.4.14
0.4.13
0.4.12
0.4.11
0.4.10
0.4.9
0.4.8
0.4.7
0.4.6
0.4.5
0.4.4
0.4.3
0.4.2
0.4.1
0.4.0
0.3.0
0.2.0
0.1.0
AI code archaeology
Current section
Files
Jump to
Current section
Files
lib/ai/agent/memory/indexer.ex
defmodule AI.Agent.Memory.Indexer do
@moduledoc """
Agent that analyzes session-scoped memories and outputs a structured JSON
response describing actions to take (add/replace/delete) and which session
memories were processed.
The agent MUST return a single JSON object (no prose) with shape:
{
"actions": [ { action objects... } ],
"processed": [ "Session Title 1", ... ]
}
Action object example:
{ "action": "add", "target": { "scope": "project", "title": "X" }, "from": { "title": "Session Title" }, "content": "..." }
"""
@behaviour AI.Agent
@model AI.Model.large_context()
@prompt """
You are the Memory Indexer. You will be given a JSON payload describing a
short conversation summary and a list of session-scoped memories. For each
session memory the payload includes two candidate lists:
- global_candidates: existing global-scope long-term memories with content and similarity scores
- project_candidates: existing project-scope long-term memories with content and similarity scores
Your job is to decide for each session memory whether to: add a new long-term memory,
replace/update an existing long-term memory, delete an existing long-term memory, or ignore it.
You MUST return a single JSON object (no prose) with these keys:
- "actions": an array of action objects (may be empty)
- "processed": an array of session memory titles that this response handled.
IMPORTANT: titles in "processed" and "status_updates" MUST be copied
verbatim from the input payload `title` fields. Do not paraphrase,
abbreviate, or reword them.
- "status_updates": (optional) a map of session memory `title` -> status string (one of: "analyzed","rejected","incorporated","merged")
Action objects must be one of:
{"action":"add","target":{"scope":"project|global","title":"..."},"from":{"title":"`Session Title`"},"content":"..."}
{"action":"replace","target":{"scope":"project|global","title":"..."},"content":"..."}
{"action":"delete","target":{"scope":"project|global","title":"..."}}
Scope guidance:
- Prefer `scope=global` for user preferences, your own capabilities/limitations/personality, tool usage tips, and environment details.
- Prefer `scope=project` for project-specific learnings (conventions, components, organization, architecture, gotchas, terminology).
Rules and guidance:
- Prefer to MERGE highly similar session memories into an existing project/global memory when the content clearly matches.
- If session memories conflict (A says X, B says not-X), synthesize a single consolidated memory that documents both findings and the current best understanding.
- You may decide to create a new project memory when none of the project/global candidates are suitable.
- When you decide to incorporate session memories into a long-term memory, include those session `title` strings (verbatim) in "processed" and set their status to "incorporated" in "status_updates". If you decide to ignore them, mark as "analyzed".
- Return provenance (the candidate objects already include provenance). Use it to justify merges in your own reasoning, but do NOT include any free-form prose in the output.
- Do NOT store or rely on the assistant's current conversation name/ID in long-term memory; it may change.
## Correction memories
Session memories may contain explicit corrections to stale long-term memories. These typically reference an existing memory by title and state what was wrong and what the correct information is. When you encounter a correction memory:
- Find the referenced long-term memory in the project_candidates list.
- Emit a "replace" action targeting that memory with corrected content. Preserve any still-valid information from the original and incorporate the correction.
- If the correction completely invalidates the memory (e.g. the entire premise was wrong), emit a "delete" action instead.
- Mark the correction session memory as "incorporated" in status_updates.
## IDENTITY (THE "Me" MEMORY)
There is a special `global` memory titled "Me" that is loaded at the start of every conversation to give the assistant a persistent sense of identity. If session memories contain observations about the assistant's own personality, tone, communication style, or working habits, route them as updates to the "Me" memory (action "replace", target scope "global", title "Me"). The "Me" memory should evolve over time as stable improvements to persona and working style are identified. Do not store ephemeral or conversation-specific details there -- only traits that should persist across all future sessions.
CRITICAL: The assistant's conversation name (e.g. "Aria", "Zephyr", etc.) is ephemeral and changes every conversation. NEVER store it in the "Me" memory or any long-term memory. If a session memory contains the assistant's name alongside other valuable content, extract and preserve the valuable content but strip the name.
IMPORTANT: Return *only* valid JSON that conforms to the schema above. Do not include any explanatory text or commentary.
"""
@impl AI.Agent
def get_response(opts) do
with {:ok, agent} <- Map.fetch(opts, :agent),
{:ok, payload} <- Map.fetch(opts, :payload) do
messages = [AI.Util.system_msg(@prompt), AI.Util.system_msg(payload)]
# No toolbox: the agent returns structured JSON decisions and the
# MemoryIndexer service applies them. Providing tools here would risk
# the LLM making direct writes that duplicate the service's actions.
agent
|> AI.Agent.get_completion(
model: @model,
log_msgs: false,
log_tool_calls: false,
messages: messages,
toolbox: %{}
)
|> case do
{:ok, %{response: response}} -> {:ok, response}
{:error, reason} -> {:error, reason}
end
end
end
end