victorvargas/rony-llm-agent
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge pull request #2 from VictorVargas/feat/consoel-buffer

Browse source 
feat: add history messages, reasoning content, and adaptive language support
This commit is contained in:
Victor Hugo Vargas Servin 2026-07-05 16:20:39 -07:00 committed by GitHub
commit 2cde90d8f7
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
8 changed files with 46 additions and 31 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
docs/architecture.es.md
View file

@ -206,8 +206,8 @@ Bucle iterativo entre LLM y ejecución de tools. Es el "cerebro" que orquesta to
```go ```go
type Loop interface { type Loop interface {
Run(ctx context.Context, input string) (Response, error) Run(ctx context.Context, input string, history ...Message) (Response, error)
RunStream(ctx context.Context, input string) iter.Seq2[Chunk, error] RunStream(ctx context.Context, input string, history ...Message) iter.Seq2[Chunk, error]
} }
type Config struct { type Config struct {

4
docs/architecture.md
View file

@ -206,8 +206,8 @@ Iterative loop between LLM and tool execution. It's the "brain" that orchestrate
```go ```go
type Loop interface { type Loop interface {
Run(ctx context.Context, input string) (Response, error) Run(ctx context.Context, input string, history ...Message) (Response, error)
RunStream(ctx context.Context, input string) iter.Seq2[Chunk, error] RunStream(ctx context.Context, input string, history ...Message) iter.Seq2[Chunk, error]
} }
type Config struct { type Config struct {

4
pkg/agent/README.es.md
View file

@ -23,8 +23,8 @@ while iteration < MaxIterations:
```go ```go
type Loop interface { type Loop interface {
Run(ctx context.Context, input string) (Response, error) Run(ctx context.Context, input string, history ...llm.Message) (Response, error)
RunStream(ctx context.Context, input string) iter.Seq2[Chunk, error] RunStream(ctx context.Context, input string, history ...llm.Message) iter.Seq2[Chunk, error]
} }
type Config struct { type Config struct {

4
pkg/agent/README.md
View file

@ -20,8 +20,8 @@ while iteration < MaxIterations:
```go ```go
type Loop interface { type Loop interface {
Run(ctx context.Context, input string) (Response, error) Run(ctx context.Context, input string, history ...llm.Message) (Response, error)
RunStream(ctx context.Context, input string) iter.Seq2[Chunk, error] RunStream(ctx context.Context, input string, history ...llm.Message) iter.Seq2[Chunk, error]
} }
type Config struct { type Config struct {

25
pkg/agent/loop.go
View file

@ -75,10 +75,12 @@ func New(cfg Config) *Loop {
} }
// Run executes the agent loop and returns the final response. // Run executes the agent loop and returns the final response.
func (l *Loop) Run(ctx context.Context, input string) (Response, error) { // Optional history messages are appended after the system prompt and before
// the new user input.
func (l *Loop) Run(ctx context.Context, input string, history ...llm.Message) (Response, error) {
start := time.Now() start := time.Now()
messages := l.buildInitialMessages(input) messages := l.buildInitialMessages(input, history)
var finalContent string var finalContent string
var allToolCalls []llm.ToolCall var allToolCalls []llm.ToolCall
var totalUsage llm.TokenUsage var totalUsage llm.TokenUsage
@ -137,9 +139,11 @@ func (l *Loop) Run(ctx context.Context, input string) (Response, error) {
} }
// RunStream executes the agent loop with streaming output. // RunStream executes the agent loop with streaming output.
func (l *Loop) RunStream(ctx context.Context, input string) iter.Seq2[llm.StreamChunk, error] { // Optional history messages are appended after the system prompt and before
// the new user input.
func (l *Loop) RunStream(ctx context.Context, input string, history ...llm.Message) iter.Seq2[llm.StreamChunk, error] {
return func(yield func(llm.StreamChunk, error) bool) { return func(yield func(llm.StreamChunk, error) bool) {
messages := l.buildInitialMessages(input) messages := l.buildInitialMessages(input, history)
iterations := 0 iterations := 0
for iterations < l.cfg.MaxIters { for iterations < l.cfg.MaxIters {
@ -177,7 +181,7 @@ func (l *Loop) RunStream(ctx context.Context, input string) iter.Seq2[llm.Stream
} }
} }
if !hasToolCalls && chunk.Delta != "" { if !hasToolCalls && (chunk.Delta != "" || chunk.ReasoningDelta != "") {
responseBuilder.WriteString(chunk.Delta) responseBuilder.WriteString(chunk.Delta)
if !yield(chunk, nil) { if !yield(chunk, nil) {
return return
@ -194,12 +198,13 @@ func (l *Loop) RunStream(ctx context.Context, input string) iter.Seq2[llm.Stream
} }
} }
func (l *Loop) buildInitialMessages(input string) []llm.Message { func (l *Loop) buildInitialMessages(input string, history []llm.Message) []llm.Message {
systemPrompt := persona.AssembleSystemPrompt(l.cfg.Persona, "") systemPrompt := persona.AssembleSystemPrompt(l.cfg.Persona, "")
return []llm.Message{ messages := make([]llm.Message, 0, len(history)+2)
{Role: llm.RoleSystem, Content: systemPrompt}, messages = append(messages, llm.Message{Role: llm.RoleSystem, Content: systemPrompt})
{Role: llm.RoleUser, Content: input}, messages = append(messages, history...)
} messages = append(messages, llm.Message{Role: llm.RoleUser, Content: input})
return messages
} }
func (l *Loop) getToolSchemas() []json.RawMessage { func (l *Loop) getToolSchemas() []json.RawMessage {

18
pkg/llm/providers/llamacpp/client.go
View file

@ -124,7 +124,8 @@ func (c *Client) Stream(ctx context.Context, req llm.CompletionRequest) iter.Seq
for _, choice := range event.Choices { for _, choice := range event.Choices {
chunk := llm.StreamChunk{ chunk := llm.StreamChunk{
Delta: choice.Delta.Content, Delta: choice.Delta.Content,
ReasoningDelta: choice.Delta.ReasoningContent,
} }
if choice.FinishReason != "" { if choice.FinishReason != "" {
chunk.FinishReason = choice.FinishReason chunk.FinishReason = choice.FinishReason
@ -210,6 +211,7 @@ func (c *Client) toResponse(resp llamaChatResponse) llm.CompletionResponse {
ID: resp.ID, ID: resp.ID,
Model: resp.Model, Model: resp.Model,
Content: choice.Message.Content, Content: choice.Message.Content,
Reasoning: choice.Message.ReasoningContent,
StopReason: choice.FinishReason, StopReason: choice.FinishReason,
} }
@ -270,9 +272,10 @@ type llamaChoice struct {
} }
type llamaMessageResult struct { type llamaMessageResult struct {
Role string `json:"role"` Role string `json:"role"`
Content string `json:"content"` Content string `json:"content"`
ToolCalls []llamaToolCall `json:"tool_calls"` ReasoningContent string `json:"reasoning_content"`
ToolCalls []llamaToolCall `json:"tool_calls"`
} }
type llamaToolCall struct { type llamaToolCall struct {
@ -306,9 +309,10 @@ type llamaStreamChoice struct {
} }
type llamaStreamDelta struct { type llamaStreamDelta struct {
Content string `json:"content"` Content string `json:"content"`
Role string `json:"role"` ReasoningContent string `json:"reasoning_content"`
ToolCalls []llamaStreamToolCall `json:"tool_calls"` Role string `json:"role"`
ToolCalls []llamaStreamToolCall `json:"tool_calls"`
} }
type llamaStreamToolCall struct { type llamaStreamToolCall struct {

10
pkg/llm/types.go
View file

@ -82,6 +82,7 @@ type CompletionResponse struct {
ID string `json:"id"` ID string `json:"id"`
Model string `json:"model"` Model string `json:"model"`
Content string `json:"content"` Content string `json:"content"`
Reasoning string `json:"reasoning,omitempty"`
ToolCalls []ToolCall `json:"tool_calls,omitempty"` ToolCalls []ToolCall `json:"tool_calls,omitempty"`
StopReason string `json:"stop_reason"` StopReason string `json:"stop_reason"`
Usage TokenUsage `json:"usage,omitempty"` Usage TokenUsage `json:"usage,omitempty"`
@ -105,10 +106,11 @@ type ToolCall struct {
// StreamChunk is emitted by the iterator returned from Stream(). // StreamChunk is emitted by the iterator returned from Stream().
type StreamChunk struct { type StreamChunk struct {
Delta string `json:"delta"` Delta string `json:"delta"`
ToolCalls []ToolCall `json:"tool_calls,omitempty"` ReasoningDelta string `json:"reasoning_delta,omitempty"`
FinishReason string `json:"finish_reason,omitempty"` ToolCalls []ToolCall `json:"tool_calls,omitempty"`
Usage TokenUsage `json:"usage,omitempty"` // only present on final chunk FinishReason string `json:"finish_reason,omitempty"`
Usage TokenUsage `json:"usage,omitempty"` // only present on final chunk
} }
// LLMClient is the interface that every provider implements. // LLMClient is the interface that every provider implements.

8
pkg/persona/persona.go
View file

@ -37,7 +37,7 @@ func DefaultPersona() Persona {
Name: "Rony", Name: "Rony",
Tone: "professional and helpful", Tone: "professional and helpful",
Style: "clear and concise", Style: "clear and concise",
Language: "en", Language: "the user's language",
} }
} }
@ -59,7 +59,11 @@ func AssembleSystemPrompt(p Persona, agentsMD string) string {
parts = append(parts, fmt.Sprintf("Write in a %s style.", p.Style)) parts = append(parts, fmt.Sprintf("Write in a %s style.", p.Style))
} }
if p.Language != "" { if p.Language != "" {
parts = append(parts, fmt.Sprintf("Respond in %s.", p.Language)) if strings.EqualFold(p.Language, "the user's language") {
parts = append(parts, "Respond in the same language as the user's messages.")
} else {
parts = append(parts, fmt.Sprintf("Respond in %s.", p.Language))
}
} }
for _, c := range p.Constraints { for _, c := range p.Constraints {
parts = append(parts, fmt.Sprintf("CONSTRAINT: %s", c)) parts = append(parts, fmt.Sprintf("CONSTRAINT: %s", c))